@blamejs/exceptd-skills 0.19.14 → 0.19.15

package/data/zeroday-lessons.json

@@ -16257,7 +16257,31 @@
     },
     "ai_discovered_zeroday": false,
     "ai_discovery_source": "vendor_research",
-    "ai_assist_factor": "none"
+    "ai_assist_factor": "none",
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-135",
+        "name": "CONTROL-PANEL-PLUGIN-USER-SURFACE-MUST-NOT-REACH-ROOT-OPERATIONS",
+        "description": "FreePBX Endpoint Manager is the control-panel module this control governs, and the packet's vector is the forbidden pattern in a single line: a testconnection feature in the administration UI reaches check_ssh_connect(), which composes an OS command from caller-supplied input (CWE-78), so a form field on a user-facing surface is the only thing standing between a panel operator and command execution on the telephony server as the asterisk user. Applied to this product, the requirement is that the connection-test path pass its parameters to the SSH invocation as arguments rather than as a shell string, and that any privileged operation the module needs sit behind its own validated interface, so the module's own input handling is not the single boundary. The stake is not reduced by the outcome being the asterisk account rather than root: asterisk is the service identity the PBX runs as, so the packet's stated outcome — remote access to the system as an asterisk user — is control of the telephony service and everything it holds, not a low-privilege shell. Distinguishing test: on a staging FreePBX, submit an Endpoint Manager connection test whose host or credential field carries shell metacharacters and confirm no subcommand executes — a patch attestation taken against the FreePBX core that never enumerates installed modules reads clean while an unpatched Endpoint Manager keeps the sink. Precondition, and this is where the control is most likely to be over-claimed on this entry: the packet's detailed vector requires an authenticated known user, but the packet's own summary of the same flaw describes unauthenticated remote command execution, so restricting who holds a FreePBX administrative account bounds the population that can reach testconnection only if the authenticated precondition is the true one — it does nothing if the summary is. Constrain the module's reachability rather than the account list until the vendor update lands, and note that this control states the property the vendor fix must establish; it does not implement it.",
+        "evidence": "Packet: CWE-78, CVSS 9.8, RWEP 77, poc_available true, active_exploitation 'confirmed', cisa_kev true with kev_date 2026-02-03. Vector: 'Sangoma FreePBX Endpoint Manager contains an OS command injection vulnerability that could allow for a post-authentication command injection by an authenticated known user via the testconnection -> check_ssh_connect() function. An attacker can leverage this vulnerability to potentially obtain remote access to the system as an asterisk user.' The packet's attack_vector field summarises the same entry as 'an OS command-injection flaw (CWE-78) enabling unauthenticated remote command execution on the telephony server' — the two disagree on whether authentication is required, and this description does not resolve that disagreement in the operator's favour. patch_available true, live_patch_available false.",
+        "gap_closes": [
+          "NIST-800-53-AC-6",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-001",
+        "name": "CISA-KEV-RESPONSE-SLA",
+        "description": "On a PBX the usual patch-window argument inverts. The packet records patch_available true, live_patch_available false, and a vendor patch that typically requires a service restart or system reboot per the KEV requiredAction — and a service restart on a production telephony server drops every call in progress, which is the specific reason this remediation gets marked done when the module update is installed rather than when the service comes back. Per the packet's own live_patch_notes, a host that has taken the update without the restart is still running the vulnerable code, so 'updated' and 'remediated' are different states here and only the second one counts. Applied to this entry, the clock starts at the 2026-02-03 KEV listing, the maintenance window is scheduled as part of the response rather than deferred into the next telephony change window, and completion is measured per host by the Endpoint Manager version actually loaded after restart rather than by an 'update applied' flag in a console. The compensating-control branch covers the interval: restrict which network segments can reach the FreePBX administration interface at all, since with a public PoC and confirmed exploitation the reachability of that interface is the remaining precondition an operator can act on. State its limits plainly — that bounds who can send the request, it does not close the injection, and it is unavailable wherever the administration interface must stay reachable for normal operation. It also does nothing for a host already reached: the packet marks exploitation confirmed, and an attacker who obtained the asterisk service identity survives the module update, so a server that was reachable during the exposure window needs its stored telephony credentials treated as disclosed and rotated rather than closed out on the patch.",
+        "evidence": "Packet: cisa_kev true, kev_date 2026-02-03, active_exploitation 'confirmed', poc_available true, CVSS 9.8, RWEP 77, ai_discovered false. patch_available true, live_patch_available false; live_patch_notes: 'No live-patch tool registered for this entry at bulk-import time. Vendor patch typically requires service restart or system reboot per the KEV requiredAction.' Vector states the outcome as an attacker potentially obtaining 'remote access to the system as an asterisk user' via the Endpoint Manager testconnection -> check_ssh_connect() function.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2",
+          "NIS2-Art21-patch-management"
+        ]
+      }
+    ]
   },
   "CVE-2019-19006": {
     "name": " Sangoma FreePBX Improper Authentication Vulnerability",
@@ -17846,7 +17870,40 @@
     },
     "ai_discovered_zeroday": false,
     "ai_discovery_source": "vendor_research",
-    "ai_assist_factor": "none"
+    "ai_assist_factor": "none",
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-134",
+        "name": "DEVICE-MANAGEMENT-GATEWAY-CONFIG-ENDPOINT-AUTH-INPUT-NEUTRALIZATION",
+        "description": "HPE OneView is the management plane for the server infrastructure it administers, and the packet places the defect on a path a remote unauthenticated user can reach: code injection (CWE-94) ending in remote code execution on the appliance itself. Bound to this product, the control means every endpoint the OneView appliance exposes authorizes its caller before the request reaches anything that evaluates content, and that caller-supplied content is neutralized rather than passed into an interpreter — and that no OneView instance is left answering from a segment with no operational need to manage infrastructure. This is also why the least-privilege gap cited on this entry does not close the path: the attacker never holds a OneView operator account, so per-account privilege scoping is never consulted and an access-review attestation passes cleanly while the injection path stays open. Distinguishing test: from a segment with no operational need to reach the management plane, send unauthenticated requests to each endpoint on a staging OneView appliance and confirm each is refused before any supplied content is evaluated. Precondition: the endpoint-side authorization and input neutralization are properties the vendor update establishes — this control states what to verify, it does not implement it. Until that update and its restart land, restricting which segments can reach the appliance bounds who can send the request but leaves it fully exploitable from anything inside the permitted segment, and that lever is unavailable for the management network the appliance must keep answering in order to administer the hardware it exists to manage.",
+        "evidence": "Packet: HPE OneView contains a code injection vulnerability (CWE-94) that allows a remote unauthenticated user to perform remote code execution on the infrastructure-management appliance; CVSS 9.8, RWEP 77; poc_available true; CISA KEV-listed 2026-01-07 with active_exploitation confirmed; NIST-800-53-AC-6 (Least Privilege) is one of the framework controls this entry cites as insufficient; patch_available true with live_patch_available false and live_patch_notes recording that the vendor patch typically requires a service restart or system reboot per the KEV requiredAction.",
+        "gap_closes": [
+          "NIS2-Art21-network-security",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-037",
+        "name": "FLEET-COMPROMISE-IR-PLAYBOOK",
+        "description": "OneView is a fleet control plane for server infrastructure, and the packet gives a pre-auth RCE with a public PoC and confirmed in-the-wild exploitation — so an appliance that was reachable between the 2026-01-07 KEV listing and its restart has to be triaged as possibly already compromised, because the vendor update closes the injection path and removes nothing an attacker executed, placed or read through it beforehand. For this CVE the playbook means: rotating every credential the appliance holds or brokers for the hardware it manages and every account that authenticated through it during that window; auditing the appliance's configuration and profile changes across the window; and having stated quarantine criteria for managed hardware whose configuration OneView could have altered while under attacker control. Precondition: this is a response control, not a preventive one — it does not lower the chance of exploitation, it bounds what an already-executed compromise retains, and it is only worth running where the exposure window can be established. Since the flaw yields code execution on the appliance, an attacker is above the appliance's own logging, so the window has to be reconstructed from network-side and collector-side records rather than trusted from the appliance's local logs.",
+        "evidence": "Packet: remote unauthenticated code execution (CWE-94) on the HPE OneView infrastructure-management appliance; CISA KEV-listed 2026-01-07 with active_exploitation confirmed; poc_available true; CVSS 9.8, RWEP 77; patch_available true, live_patch_available false, with live_patch_notes recording that the vendor patch typically requires a service restart or system reboot per the KEV requiredAction.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIST-800-53-SI-2"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-001",
+        "name": "CISA-KEV-RESPONSE-SLA",
+        "description": "An infrastructure-management appliance is usually inventoried as the tool an estate patches its servers with rather than as an asset carrying its own patch clock, which is exactly how a KEV-listed pre-auth RCE on OneView ends up riding the next infrastructure maintenance window while the operating-system patching attestation reads clean. For this CVE the control means the appliance's own update runs on the KEV clock that opened 2026-01-07, not on the cadence of the fleet updates OneView distributes, with completion measured by the version the appliance is actually running rather than by a staged or downloaded update. The packet records no live-patch path and a fix that typically requires a service restart or system reboot, so an appliance that has taken the update but not restarted still carries the vulnerable code and must be counted as exposed — and on a management appliance that restart is the step most likely to be deferred, because it interrupts the console the estate's operators depend on. Priority follows the packet: unauthenticated remote code execution with a public PoC and confirmed exploitation, on the plane that administers the hardware underneath it.",
+        "evidence": "Packet: CISA KEV-listed 2026-01-07 with active_exploitation confirmed; poc_available true; CVSS 9.8, RWEP 77; unauthenticated remote code execution (CWE-94) on the HPE OneView management appliance; patch_available true, live_patch_available false, live_patch_notes stating the vendor patch typically requires a service restart or system reboot per the KEV requiredAction; the entry cites AU-Essential-8-Patch (Patch operating systems), ISO-27001-2022-A.8.8 and NIST-800-53-SI-2 as insufficient.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2"
+        ]
+      }
+    ]
   },
   "CVE-2023-52163": {
     "name": "Digiever DS-2105 Pro Missing Authorization Vulnerability",
@@ -23766,7 +23823,21 @@
     },
     "ai_discovered_zeroday": false,
     "ai_discovery_source": "vendor_research",
-    "ai_assist_factor": "none"
+    "ai_assist_factor": "none",
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-145",
+        "name": "ACTIVELY-EXPLOITED-OS-LPE-REMEDIATION-WINDOW-ENFORCEMENT",
+        "description": "The packet puts the attacker inside the trust boundary before the flaw is touched: exploitation requires an authenticated user in the same Windows Active Directory domain as the Session Recording server, and the outcome is escalation to the NetworkService account on that server (CWE-269). For this CVE the control means the Citrix Session Recording update is driven across every Session Recording server on the clock that opened with the 2025-08-25 KEV listing rather than folded into the next Citrix maintenance window, with completion measured per server by the code actually running rather than by 'approved' or 'deployed' in a change record. The packet records a vendor patch with no live-patch path and a fix that typically requires a service restart or system reboot, so a server that has taken the update but not restarted still carries the vulnerable code and must be counted as exposed; on a recording server that restart is the step most likely to be deferred, because taking it interrupts recording of live sessions, and a deferral logged as 'patched' is the specific way this remediation goes wrong. The control's second half is the load-bearing one here: every framework control cited against this entry is a patch-timeline control, and none of them can contain an escalation whose only precondition is holding an ordinary domain account — the attacker is already authenticated, so entitlement review and account-privilege tightening leave the path fully intact while their attestations pass. Enumerate first the Session Recording servers joined to a domain whose ordinary users are not administrators of them, because there the exploit's precondition is the normal operating state rather than an anomaly. Distinguishing test: produce, per server, the running Session Recording build and the service start time, and confirm the start time is later than the update install time — a fleet report showing the fixed package installed on every server, with services still running from before the install, is a fleet that is still exploitable. Precondition: with a public PoC and confirmed in-the-wild exploitation, the update remediates the flaw but establishes nothing about whether it was used first; a server that ordinary domain accounts could reach through the exposure window needs its post-exploitation state examined rather than closed on the patch record.",
+        "evidence": "Packet fields for CVE-2024-8068 (Citrix Session Recording Improper Privilege Management Vulnerability): cwe_refs CWE-269; vector 'Citrix Session Recording contains an improper privilege management vulnerability that could allow for privilege escalation to NetworkService Account access. An attacker must be an authenticated user in the same Windows Active Directory domain as the session recording server domain.'; attack_vector records escalation of an authenticated user's privileges on the recording server. cisa_kev true with kev_date 2025-08-25; active_exploitation confirmed; poc_available true; cvss 7.8; rwep_score 77. patch_available true, live_patch_available false, live_patch_notes 'No live-patch tool registered for this entry at bulk-import time. Vendor patch typically requires service restart or system reboot per the KEV requiredAction.' Citing gaps recorded on this entry: AU-Essential-8-Patch (Patch operating systems), ISO-27001-2022-A.8.8 (Management of technical vulnerabilities), NIS2-Art21-patch-management (Vulnerability handling and disclosure), NIST-800-53-SI-2 (Flaw Remediation) and UK-CAF-B4 (System security).",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-patch-management",
+          "NIST-800-53-SI-2"
+        ]
+      }
+    ]
   },
   "CVE-2024-8069": {
     "name": "Citrix Session Recording Deserialization of Untrusted Data Vulnerability",
@@ -25232,7 +25303,39 @@
     },
     "ai_discovered_zeroday": false,
     "ai_discovery_source": "vendor_research",
-    "ai_assist_factor": "none"
+    "ai_assist_factor": "none",
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-042",
+        "name": "INCOMPLETE-PATCH-SEQUENCE-SEVERITY-MULTIPLIER",
+        "description": "The packet states outright that this CVE is a patch bypass for CVE-2025-49704 and that its updates carry more robust protection than 49704's did. That means an on-premises SharePoint farm which took the 49704 update and was recorded as remediated sat on the vulnerability-management ledger as 'patched' while remaining fully exploitable through the same CWE-502 sink — the earlier fix is the thing that failed, not the operator's SLA. Bound to this farm, the control requires that a SharePoint deserialization CVE which supersedes an earlier fix on the same sink is scored and scheduled above what its CVSS band alone implies, and that the superseded CVE's remediation record is reopened rather than left closed. It also sets the intake expectation for what follows: the packet already names a second CVE (CVE-2025-53771) this one can be chained with, so the farm's exposure is a sequence on one primitive, not a discrete ticket that closes. Precondition, and it is the limit of this control: it changes priority, scoring and record-keeping only — it remediates nothing. The vendor update plus the service restart or system reboot the packet records as typically required is what removes the code path, and a farm server that has taken the update but not the restart is still running the vulnerable code.",
+        "evidence": "Packet vector: 'CVE-2025-53770 is a patch bypass for CVE-2025-49704, and the updates for CVE-2025-53770 include more robust protection than those for CVE-2025-49704' and 'This vulnerability could be chained with CVE-2025-53771'. cwe_refs CWE-502; cisa_kev true, kev_date 2025-07-20; active_exploitation confirmed; rwep_score 83, cvss 9.8; poc_available true; patch_available true; live_patch_available false with live_patch_notes 'Vendor patch typically requires service restart or system reboot per the KEV requiredAction.'",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-vulnerability-handling"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-032",
+        "name": "PERIMETER-COMPROMISE-REBUILD-NOT-PATCH",
+        "description": "The packet's attack path does not end at code execution — it ends at web-shell deployment on an on-premises SharePoint Server reached over the network by an attacker holding no credential. For this CVE that makes the farm an already-compromised host to be triaged, not a patch target to be closed. Applied to a SharePoint deployment: every front-end and application server that was network-reachable across the exposure window is examined for attacker-written content in the directories the web application serves, and every secret the farm's application identity holds — service accounts, application credentials, and the farm's own cryptographic material — is rotated, because the update closes the deserialization sink but removes nothing already written through it or read out of it. This is also the specific reason the anti-malware control cited against this entry cannot carry the remediation: a bespoke .aspx file dropped through the application's own request path is not a signature-matched sample, so a fully deployed and current anti-malware estate reports clean over a live web shell. Preconditions: rebuild-and-rotate reaches only the farm's own hosts and its own credentials — if the foothold was used to move to an identity outside the farm, that account is out of scope here and needs separate containment; and the triage is bounded by an exposure window the operator can actually establish, so a farm without retained per-request access logs for the period cannot scope the hunt at all, in which case rebuild is the safe default rather than an evidence-driven choice.",
+        "evidence": "Packet vector: 'Microsoft SharePoint Server on-premises contains a deserialization of untrusted data vulnerability that could allow an unauthorized attacker to execute code over a network.' Packet attack_vector: 'deserialization of untrusted data (CWE-502) on SharePoint Server (the ToolShell chain), yielding unauthenticated remote code execution and web-shell deployment. CISA KEV-listed 2025-07-20 with confirmed in-the-wild exploitation.' active_exploitation confirmed; poc_available true; patch_available true; citing gap CIS-Controls-v8-10.1 'Deploy and Maintain Anti-Malware Software'.",
+        "gap_closes": [
+          "CIS-Controls-v8-10.1",
+          "NIST-800-53-SI-2"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-001",
+        "name": "CISA-KEV-RESPONSE-SLA",
+        "description": "An unauthenticated network RCE on an on-premises SharePoint farm, KEV-listed 2025-07-20 with confirmed in-the-wild exploitation, a public PoC and an RWEP of 83, is the case a KEV clock exists for — and the clock has to run to the completed restart rather than to the installed update. The packet registers no live-patch path and states the vendor patch typically requires a service restart or system reboot per the KEV requiredAction, so a farm with more than one server is remediated only when every server has both taken the update and been restarted; completion is measured per server against the fixed build, not from a deployment console reporting 'approved' or 'applied'. That distinction is where this remediation goes wrong in practice, because the restart is the step a farm operator defers to avoid taking the site offline, and a deferral recorded as patched is indistinguishable on the dashboard from a real fix. Precondition: this is a deployment clock and nothing more. It says nothing about whether the farm was exploited before the clock started — which, for a KEV-listed pre-auth RCE with confirmed exploitation and web-shell deployment, is the likely case — so meeting the SLA is not a statement that the farm is clean.",
+        "evidence": "Packet: cisa_kev true, kev_date 2025-07-20; active_exploitation confirmed; rwep_score 83; cvss 9.8; poc_available true; patch_available true; live_patch_available false; live_patch_notes 'No live-patch tool registered for this entry at bulk-import time. Vendor patch typically requires service restart or system reboot per the KEV requiredAction.' Vector: 'could allow an unauthorized attacker to execute code over a network.'",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "UK-CAF-B4"
+        ]
+      }
+    ]
   },
   "CVE-2025-25257": {
     "name": "Fortinet FortiWeb SQL Injection Vulnerability",
@@ -26329,7 +26432,30 @@
     },
     "ai_discovered_zeroday": false,
     "ai_discovery_source": "vendor_research",
-    "ai_assist_factor": "none"
+    "ai_assist_factor": "none",
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-127",
+        "name": "UNSUPPORTED-NETWORK-DEVICE-ENUMERATION-AND-RETIREMENT",
+        "description": "This packet carries two facts that have to be held together: patch_available is recorded true, and the vector states that all associated hardware revisions of the DIR-859 have reached end-of-life or end-of-service and should be retired and replaced per vendor instructions. Both are real, and together they settle the terminal state — whatever fixed firmware exists is an interim step on hardware with no ongoing patch path, so removal and replacement is the end state for every unit, not an option to weigh against patching. The requirement here is therefore an inventory that names each DIR-859 in service with its location and hardware revision and carries a dated replacement commitment, rather than a firmware-version column that turns green and stays green. A remediation record that ends at 'every DIR-859 reports the fixed build' marks the estate compliant while each unit remains exposed to everything found in that platform after the build shipped, with no vendor left to fix it — and the packet's confirmed exploitation and public PoC mean this device class is under active attention now, not hypothetically. Precondition on the interim measure, which is where this control is usually over-claimed: restricting who can reach the device's HTTP interface bounds who can send the POST, but the packet places the flaw in the router's own HTTP POST request handler, and a router exists to serve the clients behind it. Every host on the network it serves can reach that interface, so for a unit serving a general user or guest network there is no segment that removes the path — only isolation of that served network from anything else that matters, and replacement on the dated schedule.",
+        "evidence": "Packet vector: 'D-Link DIR-859 routers contain a path traversal vulnerability in the file /hedwig.cgi of the component HTTP POST Request Handler... This vulnerability affects legacy D-Link products. All associated hardware revisions have reached their end-of-life (EOL) or end-of-service (EOS) life cycle and should be retired and replaced per vendor instructions.' patch_available true; live_patch_available false; cisa_kev true, kev_date 2025-06-25; active_exploitation confirmed; poc_available true; rwep_score 77; cvss 7.8; cwe_refs CWE-22.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-vulnerability-management"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-032",
+        "name": "PERIMETER-COMPROMISE-REBUILD-NOT-PATCH",
+        "description": "The primitive the packet describes is a read, not a write: manipulating the service argument with ../../../../htdocs/webinc/getcfg/DHCPS6.BRIDGE-1.xml leaks session data from the device configuration, which the packet says potentially enables privilege escalation and unauthorized control of the device. Everything that leaked stays valid after firmware is applied — the update stops further reads, it does not invalidate what an attacker already holds, and it does not undo configuration an attacker with device control has already changed. Bound to this router: a DIR-859 that was reachable during the exposure window is treated as having had its configuration read, so the device administrative credential, the wireless keys, and any WAN or dynamic-DNS credential the configuration carries are rotated, and the unit is re-provisioned from a known-good configuration rather than left running the settings that were exposed. Preconditions, both load-bearing. First, re-provisioning is not the terminal state for this device: the packet's own vector directs retirement and replacement, so the rebuilt unit is a bridge to its replacement and any credential rotated onto it is rotated again at decommission — a rebuild recorded as closure leaves an unsupported router in service. Second, the rotation scope is only as good as the exposure window the operator can establish; a consumer-class router that retains no request logs cannot bound it, and in that case every deployed unit's configuration is treated as read.",
+        "evidence": "Packet vector: 'Manipulation of the argument service with the input ../../../../htdocs/webinc/getcfg/DHCPS6.BRIDGE-1.xml allows for the leakage of session data potentially enabling privilege escalation and unauthorized control of the device.' Packet attack_vector: 'letting an attacker read sensitive files including credentials from the device configuration. CISA KEV-listed 2025-06-25 with confirmed in-the-wild exploitation.' active_exploitation confirmed; poc_available true; patch_available true; vector's retirement-and-replacement instruction for all hardware revisions.",
+        "gap_closes": [
+          "NIST-800-53-SI-2",
+          "UK-CAF-B4"
+        ]
+      }
+    ]
   },
   "CVE-2024-54085": {
     "name": "AMI MegaRAC SPx Authentication Bypass by Spoofing Vulnerability",
@@ -28586,7 +28712,31 @@
     },
     "ai_discovered_zeroday": false,
     "ai_discovery_source": "vendor_research",
-    "ai_assist_factor": "none"
+    "ai_assist_factor": "none",
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-145",
+        "name": "ACTIVELY-EXPLOITED-OS-LPE-REMEDIATION-WINDOW-ENFORCEMENT",
+        "description": "The packet places this in the Windows Common Log File System driver and describes an authorized attacker elevating privileges locally through a use-after-free — the account is already legitimate, so no account model is being abused; a kernel privilege boundary is failing. For this CVE the control means the Windows update carrying the CLFS fix is driven across the affected estate on the KEV clock that opened 2025-05-13 rather than folded into the next monthly rollup, with completion measured per host by installed build against the fixed build for its SKU rather than by 'approved' or 'downloaded' in the management console. CLFS is a driver that ships with Windows and loads on every install, so there is no subset to carve out, no module to blacklist and no feature to disable as an interim measure — the update is the only remediation available, which is precisely why measuring it accurately is the whole of the control. The packet records a vendor patch with no live-patch path and a fix that typically requires a service restart or system reboot, so a host that has installed the update but not rebooted still runs the vulnerable driver and must be counted as exposed; shared and multi-user hosts are where that reboot is deferred, and a deferral recorded as 'patched' is the specific way this remediation goes wrong. Priority follows the packet rather than the CVSS band: at CVSS 7.8 this sorts below the 9.x remote flaws in a severity-ordered queue, while the packet records RWEP 77, a public PoC, confirmed exploitation, and that LPEs of this class are routinely paired with an initial-access flaw by ransomware operators — making it the containment step for a chain rather than a standalone endpoint item.",
+        "evidence": "Packet: use-after-free (CWE-416) in the Windows Common Log File System (CLFS) Driver allowing an authorized attacker to elevate privileges locally, exploited by a local foothold to escalate to SYSTEM; CISA KEV-listed 2025-05-13 with active_exploitation confirmed; poc_available true; CVSS 7.8 against RWEP 77; the packet describes CLFS as a recurring kernel-LPE target and notes LPEs of this class are routinely paired with an initial-access flaw by ransomware operators; patch_available true, live_patch_available false, live_patch_notes recording that the vendor patch typically requires a service restart or system reboot per the KEV requiredAction.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2",
+          "NIS2-Art21-vulnerability-management"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-003",
+        "name": "KERNEL-EXPLOITATION-DETECTION",
+        "description": "Every framework control cited on this entry is a patch or vulnerability-handling control, and none of them observe anything during the interval between the 2025-05-13 KEV listing and the reboot that completes the fix — which is the interval in which the packet says this flaw is being exploited. For this CVE the control means kernel-privilege-escalation telemetry on Windows endpoints keyed on the transition the packet actually describes: a process running under an already-authorized non-administrative account that ends up holding SYSTEM privileges, or spawning a SYSTEM-integrity child, without passing through a legitimate elevation path — correlated with that same process driving the CLFS driver from user mode. Alert within 60 seconds. What not to key on, because it would miss this exploit entirely: the packet describes a use-after-free that succeeds in escalating, not one that faults, so a rule watching for driver crashes or bugchecks never fires on a working exploit; and the packet records a public PoC, so a rule keyed on a named tool or a file hash misses the derived variant an operator will meet. The distinguishing test has to exercise the real transition rather than a stand-in: on a staging host, have a standard user account run a local escalation that ends with a SYSTEM-integrity process and confirm the rule alerts on the privilege transition itself. Preconditions: this detects, it does not prevent — its value is bounding dwell time on hosts that have not yet taken the reboot, and it does not make an unrebooted host remediated. And because the escalation's own outcome is SYSTEM, a successful attacker ends up above the agent that would report the alert, so telemetry has to reach an off-host collector to be a control at all.",
+        "evidence": "Packet: use-after-free (CWE-416) in the Windows CLFS driver exploited by a local foothold to escalate to SYSTEM, with the attacker described as an authorized user elevating privileges locally; CISA KEV-listed 2025-05-13 with active_exploitation confirmed; poc_available true; CVSS 7.8, RWEP 77; the packet notes LPEs of this class are routinely paired with an initial-access flaw by ransomware operators; patch_available true, live_patch_available false, live_patch_notes stating the vendor patch typically requires a service restart or system reboot per the KEV requiredAction; the framework controls this entry cites are all patch or vulnerability-handling controls.",
+        "gap_closes": [
+          "NIST-800-53-SI-2",
+          "UK-CAF-B4"
+        ]
+      }
+    ]
   },
   "CVE-2024-12450": {
     "name": "RAGFlow web_crawl Full-Read SSRF + Arbitrary File Read",
@@ -36261,7 +36411,31 @@
     },
     "ai_discovered_zeroday": false,
     "ai_discovery_source": "human_researcher",
-    "ai_assist_factor": "none"
+    "ai_assist_factor": "none",
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-122",
+        "name": "EOL-ASSET-DECOMMISSION",
+        "description": "The packet forecloses the remediation this entry's framework gaps assume: Cisco has not and will not release software updates that address this vulnerability, and the entry records no patch and no live-patch path. For the RV016, RV042, RV042G, RV082, RV320 and RV325 routers the packet names, remediation therefore means removal from service on a dated schedule, and the interim state is the vendor's own workaround — disabling the affected feature described in the Workarounds section — together with restricting which segments can reach the web-based management interface at all. Scope it to those six models: the packet ties the flaw to that management interface on those routers and gives no basis for treating other Cisco equipment as an instance of it. The precondition is sharp here and must be stated rather than glossed — segmentation and the feature workaround bound who can send the crafted HTTP request, they do not remove the defect, and the packet's stated access requirement is valid administrative credentials on the device, so an attacker holding or replaying those credentials from a permitted management segment still reaches the root-level command execution. Because active exploitation is confirmed and the outcome is root-level privileges plus access to unauthorized data, a unit that was reachable during the exposure window has to be treated as attacker-held: its configuration and every credential it stored are suspect, and re-provisioning it returns it to a state where no fix exists — which is why the terminal state is replacement rather than a hardened surviving unit. A requirement that ends at 'the workaround is applied' marks a device the vendor has permanently declined to fix as compliant while it stays exposed to this flaw and to anything found in it later.",
+        "evidence": "Packet: patch_available false; the vector states 'Cisco has not and will not release software updates that address this vulnerability' and directs administrators to disable the affected feature per the Workarounds section. live_patch_available false, with live_patch_notes recording no live-patch path for this product class and decommissioning as the remediation for end-of-life products. CISA KEV-listed 2025-03-03 (due 2025-03-24), active_exploitation confirmed, poc_available true, RWEP 83 against CVSS 6.5. Affected models named in the packet: RV016, RV042, RV042G, RV082, RV320, RV325. The documented path is a crafted HTTP request to the web-based management interface exploiting improper validation of user input (CWE-77); the packet states the attacker must hold valid administrative credentials on the affected device and that a successful exploit yields root-level privileges.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIST-800-53-SI-2",
+          "NIS2-Art21-vulnerability-management",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-038",
+        "name": "MITIGATION-ACTIVE-PATCH-PENDING-VERDICT-CLASS",
+        "description": "For this entry the control's middle verdict state is not a temporary one, because nothing is pending behind it: the packet records that no software update addressing this vulnerability will be released, so the feature-disable workaround is the only technical measure that will ever exist. Any RV-series unit still in service must therefore be recorded as 'vendor workaround active, no fix, none forthcoming' with a dated removal action attached, and never as satisfying a patch-cadence control. The distinguishing test is on the register rather than the device: pull the vulnerability-management record for each RV016, RV042, RV042G, RV082, RV320 and RV325 in service and read what it reports once the workaround is applied — if it reads 'remediated' or 'patched per SLA', the record has lost both the fact that a KEV-listed, actively-exploited flaw with a public PoC is still present in the device and the fact that the residual risk is now bounded only by whatever the workaround itself is worth. Precondition: this control changes what the record says, not what the device does. It reduces no exposure on its own; its whole value is preventing the workaround from closing the item and removing the schedule pressure that decommissioning depends on.",
+        "evidence": "Packet: patch_available false and the vector states Cisco has not and will not release software updates that address this vulnerability, offering only that 'administrators may disable the affected feature as described in the Workarounds section'. live_patch_available false. CISA KEV-listed 2025-03-03 (due 2025-03-24) with active_exploitation confirmed and poc_available true; RWEP 83, CVSS 6.5.",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-vulnerability-management"
+        ]
+      }
+    ]
   },
   "CVE-2023-34192": {
     "name": "Synacor Zimbra Collaboration Suite (ZCS) Cross-Site Scripting (XSS) Vulnerability",
@@ -36678,7 +36852,30 @@
         "adequate": false,
         "gap": "Patch-applications control assumes a known-vulnerable window between disclosure and exploitation; here the window was negative, so patch cadence alone can't be the only control."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-001",
+        "name": "CISA-KEV-RESPONSE-SLA",
+        "description": "The vulnerable code here is a Joomla extension rather than Joomla itself, and that is precisely where this entry's patch-cadence gaps fail: a site whose CMS core is current, and whose operating-system patching attests clean, can still be serving the unauthenticated Balbooa Forms attachment endpoint the packet describes. Applied to this CVE, the KEV clock that opened 2026-07-10 runs against an inventory of every Joomla site in the estate carrying the Balbooa Forms extension — including sites where the form sits on a low-traffic marketing page nobody treats as an application, which is the population that goes uncounted when the inventory unit is 'the website' rather than 'the extensions this website loads'. Completion is measured by the extension version each installation reports against the vendor's fixed release, not by the Joomla core version and not by a 'site updated' line in the change record. The packet records a vendor patch as available and no live-patch path, so the fixed extension release applied per installation is the remediation; the packet registers no restart requirement for this entry, so nothing in it supports treating this as a maintenance-window item. Priority follows the packet rather than the CVSS band alone: an unauthenticated upload requiring no user interaction, a public PoC, and confirmed in-the-wild exploitation mean the interval between listing and update is an interval in which the endpoint is being used. Scope: the packet ties this to the Balbooa Forms extension and to nothing else — it is a reason to know which sites carry that extension, not a reason to treat every Joomla extension in the estate as an instance of this flaw.",
+        "evidence": "Packet: CISA KEV-listed 2026-07-10, active_exploitation confirmed, poc_available true, CVSS 9.8, RWEP 65. patch_available true; live_patch_available false; live_patch_notes null (no restart or live-patch detail recorded for this entry). The affected product is the Joomla extension Balbooa Forms (CWE-434, unrestricted upload of file with dangerous type); the packet's documented path is an unauthenticated attacker submitting a .php file to a Balbooa Forms attachment field, which the extension stores in a public uploads folder with no file-type or authentication check, and which the attacker then requests directly to execute — the packet records the outcome as full RCE.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIST-800-53-SI-2",
+          "NIS2-Art21-vulnerability-management"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-032",
+        "name": "PERIMETER-COMPROMISE-REBUILD-NOT-PATCH",
+        "description": "The packet describes a flaw that does not stop mattering when the extension is updated: the attacker's artifact is a .php file already written into a public uploads folder, and requesting that URL executes it whether or not the upload path has since been repaired. On an internet-facing Joomla site, with exploitation confirmed in the wild and a public PoC, applying the fixed extension release is the first step and the compromise assessment is the rest — enumerate the Balbooa Forms upload directory and the site's web-served tree against a known-good baseline, rebuild the site from source and known-good content rather than trusting an in-place clean, and rotate every credential the web process could reach: the site database credentials, Joomla administrator accounts, keys held in site configuration, and any of those reused elsewhere. The trigger for this playbook has to be exposure rather than an alert, and the detection has to key on what the packet's path actually emits: a file with an executable extension appearing in the uploads directory outside any sanctioned change, and web-server log entries requesting a file under that uploads path directly. It will not arrive as an anti-malware hit — a small bespoke PHP file delivered through an endpoint whose purpose is accepting uploads has no signature to match — and it produces no failed-authentication event, because the packet's attacker never authenticates. Precondition: this control bounds the damage of an intrusion that already occurred. It prevents nothing on a site that has not yet been reached, and it is not a substitute for the vendor's fixed extension release, which is the only thing that closes the upload path itself.",
+        "evidence": "Packet: active_exploitation confirmed, poc_available true, CISA KEV-listed 2026-07-10, CVSS 9.8, RWEP 65. The documented exploitation path is an unauthenticated .php upload through a Balbooa Forms attachment field into a public uploads folder, which the attacker then requests directly to execute the code — the file's persistence is independent of the extension's (absent) file-type and authentication checks. patch_available true; live_patch_available false; live_patch_notes null. The packet's vector records the result as full RCE.",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "UK-CAF-B4"
+        ]
+      }
+    ]
   },
   "CVE-2026-48939": {
     "name": "iCagenda Unrestricted Upload of File with Dangerous Type Vulnerability",
@@ -37517,7 +37714,31 @@
         "adequate": false,
         "gap": "Boundary protection blocking outbound SMB (445) to arbitrary external hosts would have prevented NTLM hash relay/capture even where the client-side Protected View bypass succeeded."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-041",
+        "name": "PROTECTION-MECHANISM-FAILURE-CLASS-REGRESSION",
+        "description": "The mechanism this CVE defeats is Office Protected View itself: per the packet, a crafted 'file://' moniker hyperlink carrying the '!' new-window parameter makes Outlook bypass Protected View and open the linked remote document in edit mode when the recipient clicks it. That makes an attestation of the form 'Protected View is enforced by policy' simultaneously true and worthless — the policy is enforced and the document opens outside it, so a user-application-hardening review that inspects the policy finds nothing wrong. For this CVE the control means the moniker-link primitive becomes a permanent case in the estate's Protected-View / untrusted-document test battery, re-executed on every Office and Outlook update rather than exercised once while remediating this CVE, because a mechanism that has been bypassed once is a mechanism whose future updates need proving rather than assuming. Distinguishing test, keyed to the behaviour the packet documents rather than to a stand-in: send a managed workstation at the fixed build an email whose body carries a 'file://' moniker hyperlink with the '!' new-window parameter pointing at a share you control, click it, and check two observables — that the linked document opens in Protected View rather than edit mode, and that the workstation makes no authentication attempt to that share. The second observable is the one a naive test drops: the packet records NTLM hash leakage to an attacker SMB share as an outcome distinct from code execution, so an outbound authentication attempt is the signal that the primitive still works even when the document render looks contained, and a test that watches only the render will pass over it. Precondition: this is a verification control, not a mitigation. It tells an operator whether the protection class is actually restored on a given build; it restores nothing itself, and it gives nothing to a client that has not yet taken the vendor update. A battery that runs only at remediation time also cannot see a later regression, which is the entire reason for tying it to every update rather than to this ticket.",
+        "evidence": "Packet fields for CVE-2024-21413 (Microsoft Outlook Improper Input Validation Vulnerability): cwe_refs CWE-20; vector 'Microsoft Outlook Remote Code Execution Vulnerability'; attack_vector 'An attacker sends an email containing a crafted file:// moniker hyperlink with a ! new-window parameter; when the recipient clicks it, Outlook bypasses Office Protected View and opens the linked remote document in edit mode, enabling NTLM hash leakage via an attacker SMB share or remote code execution.' cisa_kev true with kev_date 2025-02-06; active_exploitation confirmed; poc_available true; cvss 9.8; rwep_score 74. patch_available true, live_patch_available false, live_patch_notes null. Citing gaps recorded on this entry: AU-Essential-8-Patch, ISO-27001-2022-A.8.8 (Management of technical vulnerabilities), NIS2-Art21-patch-management, NIST-800-53-SC-7 (Boundary Protection), NIST-800-53-SI-2 (Flaw Remediation) and UK-CAF-B4 (System security).",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-001",
+        "name": "CISA-KEV-RESPONSE-SLA",
+        "description": "What the clock governs on this entry is distribution, not availability: the packet records patch_available true against a CVE from the 2024 identifier series that CISA did not list until 2025-02-06, with exploitation confirmed and a public PoC. The remediation unit is also unusual for a KEV item — the vulnerable code is the Outlook client on every user's machine, not a server or an appliance, so the SLA has to be measured as the count of Outlook installs reporting the fixed build and closed against the estate's install inventory, not against a change ticket for a single system. The packet registers no live-patch path and carries no live-patch note for this entry, so each install reaches remediation only by taking the vendor update; there is no interim state in which the client is running fixed code. Precondition on the compensating-control branch, which is where this gets over-claimed: for the window before the update lands, the outbound leg the packet names — NTLM authentication to an attacker-controlled SMB share — can be restricted at the boundary, and that is what the entry's boundary-protection gap points at, but it bounds only the credential-leak half. The packet records remote code execution as a separate outcome that this restriction does not address, and it does nothing where the link's destination is reachable inside the estate rather than across the boundary being filtered, so it is a bound on one outcome for one class of destination, not a substitute for the update. Second precondition: the update closes the client path forward and recovers nothing backward. Credential material already leaked during the exposure window stays valid after every client is patched, so any account whose workstation authenticated outward to an untrusted host in that window is a rotation item and a hunt item, not a line that clears when the install count reaches 100 percent.",
+        "evidence": "Packet fields for CVE-2024-21413: attack_vector 'An attacker sends an email containing a crafted file:// moniker hyperlink with a ! new-window parameter; when the recipient clicks it, Outlook bypasses Office Protected View and opens the linked remote document in edit mode, enabling NTLM hash leakage via an attacker SMB share or remote code execution.' cisa_kev true with kev_date 2025-02-06 against a CVE in the 2024 identifier series; active_exploitation confirmed; poc_available true; cvss 9.8; rwep_score 74; cwe_refs CWE-20. patch_available true, live_patch_available false, live_patch_notes null — no live-patch path and no live-patch note is recorded for this entry. The citing gaps are dominated by patch-timeline controls (AU-Essential-8-Patch, NIS2-Art21-patch-management, NIST-800-53-SI-2, ISO-27001-2022-A.8.8), with NIST-800-53-SC-7 (Boundary Protection) recorded alongside them.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIS2-Art21-patch-management",
+          "NIST-800-53-SI-2"
+        ]
+      }
+    ]
   },
   "CVE-2022-23748": {
     "name": "Dante Discovery Process Control Vulnerability",
@@ -39072,7 +39293,30 @@
         "adequate": false,
         "gap": "Vulnerability handling processes did not flag an internet-facing admin interface as an unacceptable exposure ahead of exploitation."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-129",
+        "name": "ENTERPRISE-ERP-MANAGEMENT-FUNCTION-AUTHENTICATION",
+        "description": "The ColdFusion Administrator is the management surface on this entry, and the packet places the defect behind it: an unauthenticated attacker sends crafted requests to the ColdFusion admin API / PMS servlet and reads arbitrary files on the server filesystem, with no user interaction required. Bound to this product, the control means each administrative endpoint on that admin API decides its own caller's authorization before performing the file operation — an unauthenticated request is refused by the endpoint itself, not by whatever fronts it — and the Administrator surface is segmented so an untrusted caller cannot present the request in the first place. The packet makes that second half unusually concrete for a KEV entry: it states that exploitation requires the admin panel be exposed to the internet, so on a deployment where the Administrator answers only from an internal or VPN-reached segment, the published path has no origin to come from. The access-control gap recorded against this entry is cited precisely because no account is authenticated anywhere on this path: the attacker holds no ColdFusion administrator credential, so per-account access rules are never consulted, and an access-control attestation showing every ColdFusion administrator named, provisioned and reviewed passes cleanly while the read succeeds. Distinguishing test: from an untrusted network segment, issue unauthenticated requests to the admin API / PMS servlet paths on a staging ColdFusion instance and confirm each is refused before any file is read. Preconditions: the endpoint-side authorization is a property the vendor update establishes — this control states what to verify, it does not implement it. Removing internet reachability bounds who can send the request, and the packet gives that reachability as the exploit's stated prerequisite, but it does not repair the access control: any caller inside the permitted segment still reaches it, so a compromised internal host, a jump box or a partner-network path satisfies the precondition in full, and the measure is unavailable where the Administrator must stay reachable for operational reasons. And because the primitive is an arbitrary file read with exploitation confirmed, an instance that was internet-reachable before remediation has to have the secrets held in the files it could serve treated as read and rotated — neither the update nor the segmentation recovers what already left.",
+        "evidence": "Packet fields for CVE-2024-20767 (Adobe ColdFusion Improper Access Control Vulnerability): cwe_refs CWE-284; vector 'ColdFusion versions 2023.6, 2021.12 and earlier are affected by an Improper Access Control vulnerability that could result in arbitrary file system read. An attacker could leverage this vulnerability to access or modify restricted files. Exploitation of this issue does not require user interaction. Exploitation of this issue requires the admin panel be exposed to the internet.'; attack_vector 'An unauthenticated attacker sends crafted requests to the ColdFusion admin API/PMS servlet to read arbitrary files on the server filesystem, exposed only because the admin panel itself was reachable from the internet.' cisa_kev true with kev_date 2024-12-16; active_exploitation confirmed; poc_available true; cvss 7.4; rwep_score 70. patch_available true, live_patch_available false, live_patch_notes null. Citing gaps recorded on this entry include ISO-27001-2022-A.5.15 (Access control) and NIST-800-53-SC-7 (Boundary Protection), alongside AU-Essential-8-Patch, NIST-800-53-SI-2, NIS2-Art21-vulnerability-management and UK-CAF-B4.",
+        "gap_closes": [
+          "ISO-27001-2022-A.5.15",
+          "NIST-800-53-SC-7"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-018",
+        "name": "SCANNER-PAPER-COMPLIANCE-TEST",
+        "description": "This entry carries two independent conditions, and a scan verdict that checks only the first is paper compliance: the packet names the affected levels (ColdFusion 2023.6, 2021.12 and earlier) and separately states that exploitation requires the admin panel be exposed to the internet. Those two checks fail in opposite directions — a host at a fixed build whose Administrator still answers from the internet keeps every subsequent admin-surface defect on the published path, and a host below the fixed build that no untrusted network can route to is not on that path at all — so a vulnerability-management program that reports ColdFusion remediation as a build number has scored the wrong condition. The operational test this entry requires runs from outside: from each untrusted origin the panel could be reachable from, request the admin API / PMS servlet paths against every ColdFusion host and record whether anything answers, and report that result alongside the build level rather than in place of it. Precondition, and it is the one that makes this test worth anything: a reachability check proves only what is unreachable from the vantage point actually used. A host that one scan origin cannot route to may still answer through a cloud load-balancer rule, a partner link, a management VPN, or a hostname the scan never resolved, so a clean result from a single origin is not an exposure verdict — and an exposure check that has never returned a positive against a host known to be published is a broken check rather than a clean estate, and must be proven against a known-reachable instance before any absence is believed. The check also has to be applied retrospectively: the primitive is an arbitrary file read and exploitation is confirmed, so a host that answered externally at any point in the exposure window has to be handled as having served its files, which no version-based verdict will ever surface.",
+        "evidence": "Packet fields for CVE-2024-20767: the vector states 'ColdFusion versions 2023.6, 2021.12 and earlier are affected by an Improper Access Control vulnerability that could result in arbitrary file system read' and, separately, 'Exploitation of this issue requires the admin panel be exposed to the internet'; attack_vector records an unauthenticated attacker reading arbitrary files via the ColdFusion admin API/PMS servlet, 'exposed only because the admin panel itself was reachable from the internet'. cisa_kev true with kev_date 2024-12-16; active_exploitation confirmed; poc_available true; cvss 7.4; rwep_score 70; patch_available true. The citing gaps this addresses are the remediation-and-handling ones recorded on the entry: AU-Essential-8-Patch (Patch operating systems), NIST-800-53-SI-2 (Flaw Remediation) and NIS2-Art21-vulnerability-management (Vulnerability handling).",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIST-800-53-SI-2",
+          "NIS2-Art21-vulnerability-management"
+        ]
+      }
+    ]
   },
   "CVE-2024-50623": {
     "name": "Cleo Multiple Products Unrestricted File Upload Vulnerability",
@@ -42146,7 +42390,31 @@
         "adequate": false,
         "gap": "Least-functionality controls that still permit the Flash browser plugin leave this client-side RCE reachable; the plugin must be removed, not merely patched."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-122",
+        "name": "EOL-ASSET-DECOMMISSION",
+        "description": "A vendor fix for this specific CVE exists — the packet's vector names the fixed Flash Player builds per platform — but the product carrying it has no ongoing patch path, so reaching a fixed build is an interim state and removing the Flash Player runtime is the terminal one. Applied here: enumerate every host that still has an Adobe Flash Player runtime installed and carry each to uninstall on a bounded, dated schedule rather than an open-ended risk acceptance. Until removal completes, block Flash content at the browser and the web gateway, because the packet's path begins with a user loading a page that serves a crafted SWF — an SWF that never reaches the runtime never reaches the ExternalInterface memory-corruption sink. Precondition, and the reason the block is a holding measure and not the fix: it covers only the browsing paths it actually mediates, and the vulnerable runtime stays installed and reachable by any SWF that gets to it through a path the policy does not sit on. Scope this to the Adobe Flash Player runtime the packet names; the packet ties the ExternalInterface sink to that product and gives no mapping into other browser plugins or media runtimes, so treating every plugin in the estate as an instance of this CVE manufactures removal work against software no evidence implicates. Distinguishing test: on a representative host, confirm the Flash runtime is absent rather than confirming the browser is configured to block it — a machine whose browser blocks Flash while the runtime remains installed still carries the vulnerable code, and a record that stops at 'this host is on the fixed build' marks a product that will receive no fix for anything found after that build as compliant.",
+        "evidence": "The packet places the flaw in the ExternalInterface ActionScript functionality of Adobe Flash Player (CWE-787, CWE-119) and describes the attack as luring a user to a page serving a malicious SWF, corrupting memory and executing arbitrary code in the Flash Player process; CVSS 8.8, RWEP 48, poc_available false, active_exploitation confirmed, CISA KEV-listed 2024-09-17. The entry cites NIST-800-53-CM-7 (Least Functionality) and AU-Essential-8-App-Hardening as insufficient, and its own framework_coverage for CM-7 records least-functionality controls that still permit the Flash browser plugin as leaving this client-side RCE reachable, with the plugin needing to be removed rather than merely patched. Its defense chain records the product as end-of-life and states that only decommissioning removes the exposure.",
+        "gap_closes": [
+          "AU-Essential-8-App-Hardening",
+          "NIST-800-53-CM-7",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-001",
+        "name": "CISA-KEV-RESPONSE-SLA",
+        "description": "The clock for this entry opened on the 2024-09-17 KEV listing, and the 'verified mitigation' it demands cannot be a version bump: the entry records the product as end-of-life with no ongoing patch path, so the mitigation that satisfies the clock is removal of the Flash Player runtime from the host, or — where removal cannot complete inside the window — a documented, time-bound block of Flash content at the browser and gateway carrying a removal date. Why this has to be stated explicitly for a 2013 CVE is visible in the packet's own dates: the vector records exploitation in the wild in February 2013 and names the builds that fixed it, and CISA still listed the flaw in 2024 with confirmed in-the-wild exploitation — which only happens against installs that never reached those builds and were never removed. A vulnerability-management programme that reads this entry as 'get to the fixed build within the SLA' closes its ticket on a runtime that is frozen at a decade-old security state. Distinguishing test: for each host the KEV clock covers, produce evidence the runtime is gone, or that the content block is in force with a dated removal plan attached, rather than evidence that an installed version number is at or above the fixed build.",
+        "evidence": "The packet records CISA KEV listing on 2024-09-17 with active_exploitation confirmed, while its vector states the flaw was exploited in the wild in February 2013 and names the builds that fix it. patch_available is true and live_patch_notes records 'No live-patching primitive for this product; the vendor update (no reboot required) is the remediation.' RWEP 48 against CVSS 8.8, poc_available false. The entry cites ISO-27001-2022-A.8.8, NIST-800-53-SI-2 and NIS2-Art21-vulnerability-management as insufficient, and its defense chain records removal — not patching — as the prevention step because the product is end-of-life.",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2",
+          "NIS2-Art21-vulnerability-management"
+        ]
+      }
+    ]
   },
   "CVE-2013-0643": {
     "name": "Adobe Flash Player Incorrect Default Permissions Vulnerability",
@@ -46227,7 +46495,40 @@
         "adequate": false,
         "gap": "Boundary protection frequently leaves the EMS FmDatabaseServer TCP/8013 reachable; exposing that service to untrusted networks is the precondition for exploitation."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-128",
+        "name": "APPSERVER-REMOTING-PROTOCOL-RESTRICTION",
+        "description": "The vulnerable surface on this product is not the EMS web console but FmDatabaseServer on TCP/8013: a non-HTTP listener that answers before any authentication and whose own input handling is the defect, which is exactly why a deployment audited on web-tier hardening and perimeter firewall rules can be fully exposed. Applied to FortiClient EMS, the control means TCP/8013 on every EMS host accepts connections only from hosts with an operational need to speak that protocol to it — the endpoint population EMS serves and the management network — enforced by a host firewall on the EMS server or a network ACL in front of it, rather than inherited from the assumption that EMS sits on an internal network. The least-privilege gap recorded against this entry is not closable on this path and should not be attempted: the packet's attacker is unauthenticated and never holds an EMS account, so per-account privilege scoping is never consulted and that attestation passes cleanly while the injection runs to SYSTEM. Distinguishing test: from a general user or server VLAN with no endpoint-registration or administrative role, open a TCP connection to 8013 on a staging EMS and confirm it is refused before the listener parses anything — an estate that can show every EMS administrator authenticates, and that the appliance is filtered at the perimeter, still hands this listener to anything that can route to it internally. Precondition: reachability restriction bounds who can send the crafted packet, it does not repair the SQL handling. Any host inside the permitted set — including a compromised endpoint that legitimately talks to EMS — still reaches the injection, and the restriction is unavailable across the part of the path where endpoints must reach the listener for the product to function. Repairing the parsing itself is what the vendor update does; the packet records that update as available and as requiring no reboot.",
+        "evidence": "Packet: an unauthenticated attacker sends specially crafted packets to the FortiClient EMS FmDatabaseServer (TCP/8013) with SQL injected into the FCTUID field (CWE-89), enabling and invoking xp_cmdshell to run arbitrary commands as SYSTEM on the EMS host. Affected versions named in the packet: FortiClientEMS 7.2.0 through 7.2.2 and FortiClientEMS 7.0.1 through 7.0.10. CISA KEV-listed 2024-03-25, active_exploitation confirmed, poc_available true, CVSS 9.8, RWEP 74. patch_available true; live_patch_available false, with live_patch_notes stating there is no live-patching primitive for this product and that the vendor update (no reboot required) is the remediation. NIST-800-53-AC-6 (Least Privilege) appears among the citing gaps for this entry, while the packet's path requires no authentication.",
+        "gap_closes": [
+          "NIST-800-53-SC-7",
+          "NIS2-Art21-network-security"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-055",
+        "name": "SECURITY-TOOL-INTEGRITY-VERIFICATION",
+        "description": "The affected asset here is a security product's own management server, and this CVE is the case the control exists for: compromise of that server yields SYSTEM on the host that administers endpoints, and the builds the packet names (FortiClientEMS 7.2.0-7.2.2 and 7.0.1-7.0.10) sit on a vendor product track that most estates patch on a product cadence rather than on the KEV cadence they apply to operating systems — which is why the flaw-remediation and technical-vulnerability-management attestations on this entry can read clean while an affected EMS stays in service. The requirement for this product is that every EMS instance is enumerated in the vulnerability-management inventory as privileged software with an operating-system-grade SLA, treated as an asset with attack surface rather than as a defense that is assumed sound, and that each instance's reported version is compared against the vendor's fixed release with the clock running from the 2024-03-25 KEV listing. The packet leaves little room for deferral: it records a vendor update as available, no live-patching primitive, and no reboot required, so remediation is a service-level update on the EMS host rather than a fleet-wide maintenance event. Precondition: this control gets the fixed code onto the host and nothing more. It says nothing about what happened during the exposure window, so on any EMS whose 8013 listener was reachable while the flaw was unpatched the update has to be paired with a compromise assessment — SYSTEM-level command execution leaves artifacts that installing the fixed build does not remove.",
+        "evidence": "Packet: the product is Fortinet FortiClient EMS, affected in versions 7.2.0 through 7.2.2 and 7.0.1 through 7.0.10; the documented chain ends in arbitrary commands executing as SYSTEM on the EMS host. CISA KEV-listed 2024-03-25, active_exploitation confirmed, poc_available true, CVSS 9.8, RWEP 74. patch_available true; live_patch_available false, with live_patch_notes recording no live-patching primitive for this product and the vendor update (no reboot required) as the remediation.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-037",
+        "name": "FLEET-COMPROMISE-IR-PLAYBOOK",
+        "description": "SYSTEM on an EMS host is control of the console that administers endpoints, so response to this CVE cannot end at the EMS server. The playbook's trigger has to be exposure rather than an alert: any EMS on an affected build (7.2.0-7.2.2, 7.0.1-7.0.10) whose TCP/8013 listener was reachable between the 2024-03-25 KEV listing and the update landing gets triaged, because the packet's path produces no failed logon and no account event at all — the attacker never authenticates, which is exactly why monitoring built around identity telemetry sees nothing. The detection that does key on the documented behaviour has three signals: connections to 8013 on the EMS host from sources outside the population that legitimately speaks to it, the database service's configuration being changed to enable xp_cmdshell, and process creation as SYSTEM parented by the EMS database service. Where any of those are present, the playbook has to cover what SYSTEM on that host reaches: the EMS database contents and any credentials held in them, the deployment and configuration actions the console issued during the window, the endpoint-facing artifacts it distributes, and rotation of every credential the EMS host held or could reach. Precondition: this is a post-exposure control. It prevents no injection and does not substitute for the vendor update the packet records as available, and its triage list only reaches the endpoints and credentials the inventory can name — anything the console administers that is missing from that inventory stays unexamined.",
+        "evidence": "Packet: the chain ends with arbitrary commands executing as SYSTEM on the EMS host, reached by an unauthenticated attacker who injects SQL into the FCTUID field on FmDatabaseServer (TCP/8013) and thereby enables and invokes xp_cmdshell. active_exploitation confirmed, poc_available true, CISA KEV-listed 2024-03-25, CVSS 9.8, RWEP 74. Affected builds per the packet: FortiClientEMS 7.2.0 through 7.2.2 and 7.0.1 through 7.0.10. patch_available true, live_patch_available false, live_patch_notes recording the vendor update (no reboot required) as the remediation.",
+        "gap_closes": [
+          "UK-CAF-C1",
+          "NIST-800-53-SI-2"
+        ]
+      }
+    ]
   },
   "CVE-2024-27198": {
     "name": "JetBrains TeamCity Authentication Bypass Vulnerability",
@@ -47408,7 +47709,30 @@
         "adequate": false,
         "gap": "System-security assurance credits the WebKit/Safari sandbox; a renderer type-confusion undermines that assumption absent additional exploit-mitigation controls."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-056",
+        "name": "MOBILE-ENDPOINT-MDM-ENFORCED-KEV-SLA",
+        "description": "The packet's fix list is nine parallel builds, not one — Safari 17.3, iOS/iPadOS 15.8.7, iOS/iPadOS 16.7.5, iOS/iPadOS 17.3, macOS Monterey 12.7.3, macOS Ventura 13.6.4, macOS Sonoma 14.3, tvOS 17.3 and visionOS 1.0.2 — so for this CVE the control means the management platform drives each enrolled device to the fixed build for the OS major that device is actually on, on the clock that opened with the 2024-01-23 KEV listing, with user deferral disabled rather than discouraged. Two properties of this entry decide whether that succeeds. First, the packet records no vendor live-patch mechanism and states remediation requires applying the fixed release and rebooting: a device that has downloaded or even installed the update but has not restarted is still running the vulnerable WebKit and must be counted as exposed. On user-carried handsets the restart is the step most often deferred, and a deferral recorded as 'updated' is the specific way this remediation goes wrong. Second, tvOS 17.3 and visionOS 1.0.2 sit in the same fix list as the phones and laptops; an update SLA scoped to iPhone, iPad and Mac leaves those two builds unaddressed while the flaw-remediation attestation reads clean. Distinguishing test: enumerate the installed build and the last-restart time per enrolled device and compare each against the fixed build for that device's own OS major — not against 'latest available' and not against the console's 'deployed' or 'approved' state, both of which report success on a host that has never restarted. Precondition: this lever reaches only enrolled devices the platform can compel, which is why it is paired here with the access-condition control rather than presented as complete coverage; a personally-owned or unenrolled device carrying organizational data is outside it entirely.",
+        "evidence": "Packet: CISA KEV-listed 2024-01-23, active_exploitation confirmed, CVSS 8.8, RWEP 60, poc_available false. patch_available true; live_patch_available false with live_patch_notes 'No vendor live-patch mechanism; remediation requires applying the fixed release and rebooting.' The packet's vector names the fixed builds as 'Safari 17.3, iOS 15.8.7 and iPadOS 15.8.7, iOS 16.7.5 and iPadOS 16.7.5, iOS 17.3 and iPadOS 17.3, macOS Monterey 12.7.3, macOS Sonoma 14.3, macOS Ventura 13.6.4, tvOS 17.3, visionOS 1.0.2' and states 'Processing maliciously crafted web content may lead to arbitrary code execution.'",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIS2-Art21-patch-management",
+          "NIST-800-53-SI-2"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-126",
+        "name": "MOBILE-OS-MINIMUM-PATCH-LEVEL-ENFORCEMENT",
+        "description": "The packet states the update 'brings that fix to devices that cannot update to the latest iOS version', so this estate provably contains hardware that will not reach the current OS major. The compliance floor for this CVE is therefore a per-major table — iOS/iPadOS 15.8.7, 16.7.5 and 17.3; macOS 12.7.3, 13.6.4 and 14.3; Safari 17.3; tvOS 17.3; visionOS 1.0.2 — not a single number. A policy keyed to 'latest iOS' marks a correctly-remediated 15.8.7 device permanently non-compliant and buries the devices that actually matter; a policy keyed only to major version passes an unpatched 17.x device. The control's requirement is that this per-major floor operate as an access condition — mail, VPN and document access refused to a device below the floor for its own major — rather than as a row on a patch-compliance report, because the packet's attack path needs only that the user open attacker-controlled web content in Safari or any WebKit-based renderer, which is ordinary daily use on exactly the devices least likely to be current. Distinguishing test: enrol a device pinned one build below the floor for its major and confirm the policy actually denies it access to protected resources; an estate that surfaces the stale build on a report while the device keeps its mail and VPN has recorded the exposure rather than removed it. Preconditions, stated rather than assumed: denying access bounds what organizational data an exploited device can reach — it does not remove the vulnerable renderer from the device, so the user's personal browsing on that handset still reaches the sink, and it does nothing for a device compromised during the exposure window, which belongs on the incident path. It is also a holding measure only until the fixed release and its required restart land on that device. And for hardware that cannot take the current major, reaching the backported build is the ceiling that hardware allows for this CVE, not a resting state: those units need a dated replacement schedule, because a floor that can only ever name backported builds marks aging hardware compliant while everything not backported accumulates against it.",
+        "evidence": "Packet: vector states 'This fix associated with the Coruna exploit was shipped in iOS 17.3 on January 22, 2024. This update brings that fix to devices that cannot update to the latest iOS version,' and lists fixed builds across Safari 17.3, iOS/iPadOS 15.8.7, 16.7.5 and 17.3, macOS Monterey 12.7.3, Sonoma 14.3, Ventura 13.6.4, tvOS 17.3 and visionOS 1.0.2. Attack vector: 'A victim opens attacker-controlled web content in Safari or any WebKit-based renderer; the content triggers a type confusion in WebKit, letting the attacker execute arbitrary code in the browser process as a foothold on the device.' CWE-843; CISA KEV-listed 2024-01-23 with active_exploitation confirmed; poc_available false, so there is no public exploit artifact for a signature to key on and prevention carries the weight; patch_available true with live_patch_available false and live_patch_notes 'No vendor live-patch mechanism; remediation requires applying the fixed release and rebooting.'",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "UK-CAF-B4"
+        ]
+      }
+    ]
   },
   "CVE-2023-34048": {
     "name": "VMware vCenter Server Out-of-Bounds Write Vulnerability",
@@ -48423,7 +48747,31 @@
         "adequate": false,
         "gap": "Application-hardening guidance focuses on macro/Office hardening on endpoints, not on server-side spreadsheet parsers that eval format strings."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-021",
+        "name": "TIER-3-DEPENDENCY-INVENTORY",
+        "description": "The vulnerable code here is a third-party Perl module, and the packet's exploitation path is 'a service embedding Spreadsheet::ParseExcel' — so an operator's exposure is a property of software it bought or built on, not of anything its own package list names. Bound to this CVE, the requirement is that the component inventory resolve to the depth at which Spreadsheet::ParseExcel actually sits — inside an appliance image, a container layer, or a vendor product that ingests spreadsheets — and record the module version, obtained from the supplier's SBOM or a direct written answer wherever the operator cannot inspect the image itself. Hold the scope there: the packet ties the string-eval sink to this module and the services embedding it, and provides no mapping from the defect into other spreadsheet-parsing code, so widening the sweep to every library in the estate that reads Excel files manufactures findings and rework against components no evidence implicates. The vulnerability-management and flaw-remediation controls cited as insufficient for this entry fail at their input rather than in their process: both act on an asset register, and a module bundled three levels down never reaches one, so the programme can be fully conformant and never open a ticket for a KEV-listed remote code execution it is running every day. Distinguishing test: for each product in the estate that ingests spreadsheets, ask whether it embeds Spreadsheet::ParseExcel and at what version, and require the answer to come from the supplier — because querying the operator's own package manager returns 'not installed' on a fully exposed system, which is the specific way this check gets recorded as clean. Precondition and limit: inventory is visibility, not remediation. The packet records a vendor fixed release as available and no live-patch mechanism, so every copy the inventory surfaces still has to be taken to that release by whoever controls it; for a bundled copy that is the embedding vendor, and until they ship it the operator's position is a known-exposed parsing path to be constrained, not a closed finding.",
+        "evidence": "Packet: 'Spreadsheet::ParseExcel version 0.65 is a Perl module used for parsing Excel files. Spreadsheet::ParseExcel is vulnerable to an arbitrary code execution (ACE) vulnerability due to passing unvalidated input from a file into a string-type \"eval\"', with the issue stemming from evaluation of Number format strings within the Excel parsing logic; attack vector records that the code executes when 'a service embedding Spreadsheet::ParseExcel parses the file'; CWE-95 and CWE-94; CISA KEV-listed 2024-01-02 with active_exploitation confirmed and poc_available true; RWEP 66 / CVSS 7.8; patch_available true and live_patch_available false, with the live-patch note recording that 'remediation requires applying the vendor fixed release'. ISO 27001:2022 A.8.8, NIST 800-53 SI-2 and DORA Art.9 are the framework controls recorded as insufficient for this entry.",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2",
+          "DORA-Art-9"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-116",
+        "name": "MULTI-TRIGGER-DEPENDENCY-EXECUTION-POLICY",
+        "description": "This CVE adds a dependency-execution trigger that no policy enumerates: document-parse time. The packet has the module passing a Number-format string out of an attacker-crafted Excel file into a Perl string eval, so the attacker's code runs when a service performs an ordinary data-processing operation on a business document — not when the dependency is installed, imported, or built. Every safeguard an estate is likely to hold is keyed to a moment that is never reached: ignore-scripts defaults and lockfile pinning govern install time, build sandboxes govern compile time, and none of them is consulted when a mail-handling service, a reporting job, or an upload handler opens the attachment. Applied here, the policy must list document/data-parse time alongside install, import and build time as an execution trigger, and every service that feeds files into Spreadsheet::ParseExcel must run under an identity scoped to that job alone — no credentials the parse itself does not need, no interactive shell, and constrained egress — so that code executed out of a format string lands somewhere that cannot reach the estate's secrets or call outward. The user-application-hardening control cited as insufficient here is aimed at a different machine entirely: it governs macro, OLE and add-in behaviour in an Office client on a user's endpoint, while this code executes inside a server-side Perl process with that service's privileges, on a file no person may ever open. Distinguishing test: feed a file whose Number-format string carries a Perl payload to a staging instance of each service that parses spreadsheets through this module, and confirm from the service account's own context that the payload can neither read credentials nor open an outbound connection — an attestation covering endpoint Office hardening passes cleanly while the server-side parse path executes attacker code unimpeded. Precondition: this constrains what the executed code can reach; it does not stop the execution. The packet records a vendor fixed release with no live-patch path, so every copy still has to be taken to that release, and the least-privilege identity is worth nothing on a service whose parse job simply inherited the application's main credentials — which is the normal case, and the thing to change first.",
+        "evidence": "Packet: the flaw stems from 'the evaluation of Number format strings (not to be confused with printf-style format strings) within the Excel parsing logic', passing unvalidated input from a file into a string-type eval; attack vector records that an attacker crafts an Excel file whose Number-format string contains Perl code and the module executes it when a service embedding Spreadsheet::ParseExcel parses the file; CWE-95 and CWE-94; KEV 2024-01-02, active_exploitation confirmed, poc_available true; patch_available true, live_patch_available false, remediation is applying the vendor fixed release. AU Essential Eight user application hardening, NIS2 Art.21 security of network and information systems, and UK CAF B4 (System security) are the framework controls recorded as insufficient for this entry.",
+        "gap_closes": [
+          "AU-Essential-8-App-Hardening",
+          "NIS2-Art21-network-security",
+          "UK-CAF-B4"
+        ]
+      }
+    ]
   },
   "CVE-2023-7024": {
     "name": "Google Chromium WebRTC Heap Buffer Overflow Vulnerability",
@@ -49747,7 +50095,32 @@
         "adequate": false,
         "gap": "Applying vendor patches within the ISM timeframe still leaves the zero-day window uncovered; the control does not mandate the EDR/exploit-guard telemetry that would catch DWM token abuse."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-145",
+        "name": "ACTIVELY-EXPLOITED-OS-LPE-REMEDIATION-WINDOW-ENFORCEMENT",
+        "description": "The packet places this in the Windows Desktop Window Manager (DWM) Core Library and describes an attacker who already holds a low-privilege foothold abusing an untrusted-pointer / memory-corruption flaw to move from the Window Manager\\DWM context the packet names to SYSTEM. The escalation starts from an account the host already accepts, so nothing in the account model is being abused — a privilege boundary inside the component is failing. For this CVE the control means driving the Windows update carrying the DWM Core Library fix across every affected host on the clock that opened with the 2023-11-14 KEV listing rather than folding it into the next monthly rollup, with completion measured per host by the installed build against the fixed build for that Windows release, not by 'approved' or 'downloaded' in the management console. The packet records no vendor live-patch mechanism and remediation requiring the fixed release plus a reboot, so a host that has installed the update but not restarted is still running the vulnerable library and has to be counted as exposed; on a shared workstation or session host that restart is the step most likely to be deferred because taking it evicts logged-on users, and a deferral recorded as 'patched' is the specific way this remediation goes wrong. Enumerate first the hosts where an unprivileged interactive logon is the normal operating state, because that is where the exploit's stated precondition — an existing low-privilege foothold — is met continuously rather than exceptionally. The control's second half is the load-bearing one here: the entry cites NIST-800-53-AC-6 as insufficient and it is, because the attacker is a local user by precondition and tightening what that account may do does not contain a flaw that hands it SYSTEM, which is why a least-privilege attestation passes cleanly while the escalation stays fully available.",
+        "evidence": "Packet: CISA KEV-listed 2023-11-14, active_exploitation confirmed, CVSS 7.8, RWEP 59, CWE-822 / CWE-119. attack_vector: 'After gaining a low-privilege foothold, an attacker abuses a memory-corruption / untrusted-pointer flaw in the DWM Core Library to escalate from the Window Manager\\DWM context to SYSTEM.' patch_available true; live_patch_available false; live_patch_notes: 'No vendor live-patch mechanism; remediation requires applying the fixed release and rebooting.' NIST-800-53-AC-6 (Least Privilege), NIST-800-53-SI-2 (Flaw Remediation), ISO-27001-2022-A.8.8, NIS2-Art21-patch-management, AU-ISM-1546 and UK-CAF-B4 are all recorded as insufficient controls on this entry.",
+        "gap_closes": [
+          "NIST-800-53-SI-2",
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-patch-management",
+          "AU-ISM-1546",
+          "NIST-800-53-AC-6",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-003",
+        "name": "KERNEL-EXPLOITATION-DETECTION",
+        "description": "Bound to this CVE, the signal to instrument is the one transition the packet actually documents: a process running in the Window Manager\\DWM context, or under an ordinary interactive user, obtaining SYSTEM on a host where no legitimate elevation path — an operator-consented UAC elevation, a service start by the Service Control Manager, a scheduled-task launch — accounts for it. Alerting keys on that token/integrity-level transition and on the SYSTEM-integrity child processes that follow it. Two other signals are the wrong answer and both are worth naming, because they are the ones a rule-writer reaches for by default. A rule keyed on an exploit artifact's hash, command line or loaded-module set cannot exist honestly here: the packet records poc_available false, so there is no public exploit to derive a signature from and any such rule would be built from an assumption. A rule keyed on dwm.exe crashing is worse than useless, because the packet describes the untrusted-pointer flaw being abused to escalate, not to fault — a working exploit produces a privilege transition and no crash, so the rule stays silent through exactly the case it was written for. Instrumentation is user-mode process and token telemetry: the packet places the flaw in the DWM Core Library, not in a kernel driver, so driver-load and kernel-callback signals do not observe this path. The distinguishing test is to force a benign SYSTEM-token acquisition from an unprivileged interactive process on a staging host and confirm the alert fires from the transition alone, with no exploit sample present. Preconditions: this detects, it does not prevent — by the time the alert exists the escalation has already happened, so it cannot be recorded as this CVE's mitigation, which the packet gives as the vendor fixed release plus a reboot. And it covers only hosts where the telemetry is deployed and reporting, a smaller population on any real estate than the one the patch report counts.",
+        "evidence": "Packet: attack_vector documents the escalation 'from the Window Manager\\DWM context to SYSTEM' following a low-privilege foothold, and locates the flaw in the DWM Core Library. poc_available is false. active_exploitation is confirmed and the entry is CISA KEV-listed 2023-11-14. NIST-800-53-SI-4 (System Monitoring) is recorded as an insufficient control on this entry. Remediation per live_patch_notes is applying the fixed release and rebooting; live_patch_available is false.",
+        "gap_closes": [
+          "NIST-800-53-SI-4"
+        ]
+      }
+    ]
   },
   "CVE-2023-36025": {
     "name": "Windows SmartScreen Security Feature Bypass",
@@ -49873,7 +50246,32 @@
         "adequate": false,
         "gap": "The 48-hour patch-exploited-flaw target still leaves the zero-day window open, and the control does not mandate the kernel/EDR telemetry that would surface driver-level escalation."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-145",
+        "name": "ACTIVELY-EXPLOITED-OS-LPE-REMEDIATION-WINDOW-ENFORCEMENT",
+        "description": "The packet puts the attacker on the host already — from a low-privilege foothold, driving crafted placeholder and reparse-point operations into the Cloud Files Mini Filter Driver to reach a heap-based buffer overflow and land as SYSTEM. The account is legitimate throughout and no boundary between users is crossed, so nothing in the account model is being abused: a privilege boundary inside a Windows driver is failing. For this CVE the control means the Windows update carrying the fix is driven across the affected estate on the clock that opened with the 2023-11-14 KEV listing rather than folded into the next monthly rollup, with completion measured by each host's installed build against the fixed build for its SKU — not by 'approved', 'downloaded' or 'deployed' in the management console. The packet records no vendor live-patch mechanism and states remediation requires applying the fixed release and rebooting, so a host that has installed the update but not restarted is still running the vulnerable driver and must be counted as exposed; a pending-reboot host reported as patched is the specific way this remediation goes wrong. The population to enumerate is the general Windows workstation estate rather than one server role: the packet names a Windows driver component and states only one precondition — a low-privilege foothold — which is the normal operating state of any machine where ordinary users hold interactive sessions. The control's second half is the load-bearing one here. Because the attacker is already an authorized local user, tightening per-account privilege does not contain the escalation, which is exactly why the least-privilege gap cited on this entry can pass a clean attestation while the flaw remains fully exploitable. Priority follows the packet rather than the 7.8 CVSS band: confirmed in-the-wild exploitation of a foothold-to-SYSTEM step makes this a chain-containment item, not a routine endpoint ticket.",
+        "evidence": "Packet: CWE-122 and CWE-787; CISA KEV-listed 2023-11-14 with active_exploitation confirmed; CVSS 7.8, RWEP 59; poc_available false. Attack vector: 'From a low-privilege foothold, an attacker triggers a heap-based buffer overflow in the Cloud Files Mini Filter Driver via crafted placeholder/reparse-point operations, escalating to SYSTEM.' patch_available true; live_patch_available false with live_patch_notes 'No vendor live-patch mechanism; remediation requires applying the fixed release and rebooting.' NIST-800-53-AC-6 (Least Privilege) is among the framework control gaps citing this CVE.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-patch-management",
+          "NIST-800-53-SI-2",
+          "NIST-800-53-AC-6",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-003",
+        "name": "KERNEL-EXPLOITATION-DETECTION",
+        "description": "The monitoring gap cited on this entry is the one the patch clock does not close, and its shape here is set by a specific packet fact: poc_available is false while active exploitation is confirmed, so there is no public exploit for a tool signature or hash to key on and only the behaviour the packet describes is available to detect. Two signals follow from it directly. First the operation: repeated crafted placeholder and reparse-point operations against the Cloud Files Mini Filter Driver issued by a process with no cloud-sync role — the processes that legitimately drive the cloud-files placeholder path on a host are its sync clients and the shell, so the same operations arriving from an unrelated user process are the anomaly, and a heap-overflow attempt against a driver is far likelier to be repeated than single-shot, which makes the repetition itself part of the signal. Second the outcome: a privilege transition with no corresponding authorized elevation — an already-running non-elevated process coming to hold a SYSTEM token, or a SYSTEM-level child spawned by a non-elevated parent — correlated back to the process that issued those driver operations, which is what turns an alert into the initial-access foothold a responder can pivot from. The control's 60-second alerting requirement is what makes this useful, because the escalation exists to precede the attacker's next action rather than to be the action. Preconditions and limits, stated because they change what this control can be recorded as achieving: it is post-hoc — it fires after SYSTEM has already been obtained, so it bounds dwell time rather than preventing the escalation, and it is not a substitute for the reboot-gated update. It also misses an exploit that uses the acquired token in-process without spawning anything, which is why the driver-operation signal must be collected in its own right and not merely as context attached to a process-spawn alert. And a host whose endpoint telemetry is not collected at kernel-object and process-token granularity produces neither signal at all, so estate coverage has to be verified on a sample rather than assumed from the agent's install count.",
+        "evidence": "Packet: poc_available false with active_exploitation confirmed and CISA KEV listing 2023-11-14 — exploitation is occurring without a public exploit artifact to signature. Attack vector: 'From a low-privilege foothold, an attacker triggers a heap-based buffer overflow in the Cloud Files Mini Filter Driver via crafted placeholder/reparse-point operations, escalating to SYSTEM' — the driver operations and the privilege transition are both stated behaviours. CWE-122 (heap-based buffer overflow) and CWE-787. NIST-800-53-SI-4 (System Monitoring) is among the framework control gaps citing this CVE. patch_available true with live_patch_available false and live_patch_notes 'No vendor live-patch mechanism; remediation requires applying the fixed release and rebooting,' so detection covers the window until each host has been restarted onto the fixed build.",
+        "gap_closes": [
+          "NIST-800-53-SI-4"
+        ]
+      }
+    ]
   },
   "CVE-2023-47246": {
     "name": "SysAid Server Path Traversal to Code Execution",
@@ -52949,7 +53347,39 @@
         "adequate": false,
         "gap": "Periodic technical-vulnerability management cannot compel the emergency out-of-band upgrade to 6.9.5 / 7.0.2 that a KEV-listed unauthenticated RCE with a 3-day remediation window demands."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-001",
+        "name": "CISA-KEV-RESPONSE-SLA",
+        "description": "The packet gives an unauthenticated path from the open internet to remote code execution on a default WordPress install, so the clock this control sets is what stands between the 2026-07-21 KEV listing and a site being taken. The remediation the packet names is upgrading to 6.9.5 or 7.0.2, and the specific failure mode for this CVE is treating the forced automatic security update WordPress.org enabled as coverage. That push is a delivery mechanism, not an inventory: completion has to be measured per site by the core version each install actually reports, so that sites with core auto-updates disabled, deployed on a read-only filesystem, or built through a pipeline that pins core are surfaced as still vulnerable instead of assumed carried by the vendor. Scope the enumeration to the WordPress installs the organisation operates — including the ones that never reached an application inventory, such as campaign microsites, documentation and event sites, and multisite networks where one core version serves many hostnames — and not to web software generally; the packet ties this defect to the WordPress REST batch endpoint and implicates nothing else. The distinguishing test: pull the reported core version from every site the organisation serves and confirm each is at or above the fixed release, rather than confirming that automatic updates are enabled in policy — a policy attestation reads clean on an install whose updater has been disabled for years. Precondition: reaching the fixed version closes the dispatch defect for requests arriving afterwards and nothing more. Exploitation is confirmed and a public exploit exists, so for a site that was internet-reachable before the upgrade landed, the version bump is the start of the response, not the end of it.",
+        "evidence": "Packet: CISA KEV-listed 2026-07-21, active_exploitation confirmed, poc_available true, CVSS 9.8, RWEP 79, CWE-436. vector: 'WordPress Core REST API batch endpoint (/wp-json/batch/v1) has an array-desynchronization interpretation conflict that misroutes handler dispatch, letting an unauthenticated attacker bypass validation to perform SQL injection and achieve remote code execution.' patch_available true; live_patch_available false; live_patch_notes: 'No live-patch mechanism; WordPress.org enabled forced automatic security updates. Remediation is upgrading to 6.9.5 / 7.0.2; interim mitigation is WAF-blocking /wp-json/batch/v1.' AU-Essential-8-Patch, NIST-800-53-SI-2, ISO-27001-2022-A.8.8 and UK-CAF-B4 are recorded as insufficient controls on this entry.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIST-800-53-SI-2",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-038",
+        "name": "MITIGATION-ACTIVE-PATCH-PENDING-VERDICT-CLASS",
+        "description": "The packet names an interim mitigation — WAF-blocking /wp-json/batch/v1 — alongside the fixed releases 6.9.5 and 7.0.2, and this control exists to stop the first being recorded as the second. A site running the WAF rule on a pre-fix core is in state (b): the batch endpoint's array-desynchronization defect is untouched and what stands between an attacker and it is a request-matching rule, whose failure reopens unauthenticated remote code execution in full. Carry that on the compliance record as its own state with a dated action item to reach the upgrade, rather than closing the finding as remediated. The preconditions that must travel with the rule, stated rather than assumed: it covers only requests that actually traverse the WAF, and only those its pattern matches. An origin reachable directly by IP or by a hostname that does not resolve through the proxy is unprotected; so is any route to the same REST handler that the pattern does not anticipate. And the reason that second precondition is sharper here than for an ordinary WAF-mitigated CVE is the defect itself — an interpretation conflict that desynchronizes handler dispatch is a disagreement about what a request means, which is precisely the assumption a request matcher in front of the handler depends on. Treat the rule as bounding exposure, never as removing the endpoint. The distinguishing test: from outside the network, send a batch-endpoint request to the site's origin address rather than through the WAF-fronted hostname, and confirm it is refused — a site whose audit record reads 'mitigated' while the origin answers directly is in state (c) and does not know it.",
+        "evidence": "Packet live_patch_notes: 'No live-patch mechanism; WordPress.org enabled forced automatic security updates. Remediation is upgrading to 6.9.5 / 7.0.2; interim mitigation is WAF-blocking /wp-json/batch/v1.' live_patch_available false; patch_available true. vector and attack_vector describe an 'interpretation conflict' / 'array-desynchronization' in the /wp-json/batch/v1 endpoint that 'misroutes handler dispatch', letting an unauthenticated attacker bypass validation. CVSS 9.8, RWEP 79, poc_available true, active_exploitation confirmed, KEV-listed 2026-07-21. NIST-800-53-SI-2 (Flaw Remediation) and ISO-27001-2022-A.8.8 (Management of technical vulnerabilities) are recorded as insufficient controls on this entry.",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-032",
+        "name": "PERIMETER-COMPROMISE-REBUILD-NOT-PATCH",
+        "description": "The packet's chain does not stop at code execution: the attacker bypasses validation, smuggles the CVE-2026-60137 SQL injection, forges an administrator account, and uploads a malicious plugin. Each of those is persistence that lives in the site's own database and filesystem, and upgrading to 6.9.5 or 7.0.2 removes none of them — a forged administrator is still a valid account after the upgrade, and an uploaded plugin is still loaded on the next request. An internet-facing WordPress install that takes a pre-auth RCE is in the same position this control was written for even though it is not a network appliance: the system that was breached is the system whose own records would have to be trusted to report the breach. So with exploitation confirmed and a public exploit available, an install that was serving before the fix landed is handled as possibly-compromised rather than closed on the version bump. Enumerate administrator and other privileged accounts with their creation times, compare the installed plugin set against a known-good manifest, and rotate what the exposed install held — database credentials, the site's authentication keys and salts, and any API keys or tokens stored in its configuration, since the SQL injection in the chain reaches the same database that holds them. Where those checks cannot be made conclusive, rebuild from a known-good baseline and re-import content instead of inheriting the running filesystem. Precondition, and the sequencing matters: this is the response for an install exposed during the window, not a substitute for the upgrade. Running the eviction work while the site is still on a pre-fix core hands the attacker the same path straight back in, so the upgrade — or, failing that, the packet's interim WAF block on /wp-json/batch/v1 with its own limits understood — goes first, and the account, plugin and credential work follows it.",
+        "evidence": "Packet attack_vector: 'An interpretation conflict in the WordPress REST batch endpoint desynchronizes handler dispatch, letting an unauthenticated attacker bypass validation, smuggle the CVE-2026-60137 SQLi, forge an administrator, and upload a malicious plugin for RCE on default installs (wp2shell).' active_exploitation confirmed; poc_available true; CISA KEV-listed 2026-07-21; CVSS 9.8, RWEP 79. patch_available true with live_patch_notes naming 6.9.5 / 7.0.2 as the remediation and WAF-blocking /wp-json/batch/v1 as the interim mitigation. NIS2-Art21-incident-handling (Incident handling) is recorded as an insufficient control on this entry.",
+        "gap_closes": [
+          "NIS2-Art21-incident-handling"
+        ]
+      }
+    ]
   },
   "CVE-2026-0770": {
     "name": "Langflow Inclusion of Functionality from Untrusted Control Sphere Vulnerability",
@@ -53010,7 +53440,42 @@
         "adequate": false,
         "gap": "Technical-vulnerability management cannot drive a patch that the vendor has not shipped; without a fixed release it does not compel the compensating network isolation and endpoint removal that actually stop exploitation of this unauthenticated exec() surface."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-103",
+        "name": "AI-APP-BUILDER-EXECUTION-ENDPOINT-AUTH-AND-SANDBOX",
+        "description": "Langflow is the visual LLM app/agent builder this control names, and CVE-2026-0770 is the precise defect it forbids: /api/v1/validate/code answers an unauthenticated POST and validate_code() hands the request body straight to exec(), with an exec_globals context that still exposes importlib and builtins, so attacker-supplied Python runs as the Langflow process — root in a default container. Bound to this product the requirement is that no route reaching that exec() sink answers an unauthenticated caller, and that any code the platform evaluates on a user's behalf runs with no filesystem, network or process reach beyond the flow's intent, instead of inside the server's own interpreter. Precondition, which is where this control is normally over-claimed: the packet records no patched version, so an operator cannot establish that property by updating. What they can do is what the packet names — block or remove exposure of /api/v1/validate/code at the reverse proxy, network-isolate the instance, and keep port 7860 off untrusted networks. That bounds who can send the POST; it does not repair the sink, it holds only for the routes the proxy actually fronts, and it is unavailable wherever the builder must stay reachable for normal use. It also does nothing for an instance already exploited: because exploitation is confirmed and the observed activity steals AWS credentials and cloud metadata, an instance that was reachable during the exposure window needs the container's cloud identity rotated and the flows and container rebuilt, not merely fronted with a new proxy rule. Distinguishing test: from an unauthenticated client against a staging instance, POST to /api/v1/validate/code a payload that imports a module and opens a network connection, and confirm it is refused before any code runs — an attestation that the AI platform is 'hardened' passes cleanly while a reachable 7860 keeps an unauthenticated exec() live.",
+        "evidence": "Packet vector: Langflow exposes an unauthenticated /api/v1/validate/code endpoint whose validate_code() passes attacker-supplied Python to exec() with an exec_globals context exposing importlib and builtins, yielding remote code execution as the Langflow process; attack_vector adds 'root in default containers' and that observed exploitation steals AWS credentials and cloud metadata. CWE-829, CVSS 9.8, RWEP 82, poc_available true, active_exploitation confirmed, CISA KEV-listed 2026-07-21. patch_available false and live_patch_available false; live_patch_notes record GHSA-g22f-v6f7-2hrh with first_patched_version: null, that ZDI published this as a zero-day, that a fixed release could not be confirmed from a primary source, that the CISA KEV required action is vendor mitigation or product discontinuation, and that the compensating controls are blocking/removing /api/v1/validate/code exposure, network-isolating Langflow, and not exposing port 7860 to untrusted networks.",
+        "gap_closes": [
+          "AU-Essential-8-App-Hardening",
+          "ISO-27001-2022-A.8.8",
+          "UK-CAF-B4",
+          "NIS2-Art21-network-security"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-038",
+        "name": "MITIGATION-ACTIVE-PATCH-PENDING-VERDICT-CLASS",
+        "description": "For this CVE the three-state distinction is not a refinement, it is the whole finding: state (a) is unreachable. The packet records patch_available false, GHSA-g22f-v6f7-2hrh carrying first_patched_version: null, and a fixed release that could not be confirmed from a primary source — so no Langflow deployment can legitimately be recorded as 'patched per SLA' for CVE-2026-0770, and a remediation register that closes this row on a patch is recording an event that has not happened. The verdict every instance must carry instead is state (b): compensating controls active, no binary patch, with the specific set named on the record — /api/v1/validate/code exposure blocked or removed, the instance network-isolated, port 7860 not reachable from untrusted networks — and the residual risk stated as what it is, that any gap in the proxy rule or any alternate route into the same exec() sink reopens unauthenticated RCE. Because the CISA KEV required action offers only vendor mitigation or product discontinuation, the time-bound action item this control demands cannot be 'apply the patch when it ships'; it has to be a dated decision point at which continuing to run Langflow is re-authorised against discontinuing it, so that an instance with no patch path does not sit under an open-ended risk acceptance while the ISMS reads clean. Distinguishing test: pull the vulnerability register's row for this CVE and confirm it shows a compensating-control state with a named review date and a discontinuation decision point, not a closed 'remediated' verdict — an SI-2 flaw-remediation attestation that treats the mitigation as a patch marks a confirmed-exploited, no-fix RCE compliant.",
+        "evidence": "Packet records patch_available false, live_patch_available false, and live_patch_notes stating GHSA-g22f-v6f7-2hrh has first_patched_version: null, that ZDI published this as a zero-day, that a fixed release could not be confirmed from a primary source, and that the CISA KEV required action is vendor mitigation or product discontinuation, with the compensating controls enumerated as blocking/removing /api/v1/validate/code exposure, network-isolating Langflow, and not exposing port 7860 to untrusted networks. active_exploitation confirmed; CISA KEV-listed 2026-07-21; RWEP 82; CVSS 9.8.",
+        "gap_closes": [
+          "NIST-800-53-SI-2",
+          "ISO-27001-2022-A.8.8"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-033",
+        "name": "AI-ML-DEVELOPER-TOOLING-INVENTORY",
+        "description": "Langflow is an agent-framework control plane of exactly the kind this inventory exists to capture, and with no patched version recorded the inventory is the precondition for every other action on this CVE: the compensating controls the packet names can only be applied to instances someone knows about. Scoped to this product, the inventory must name every Langflow deployment with its listening endpoint (port 7860 by default), whether /api/v1/validate/code is reachable and from which networks, and the blast radius — which for CVE-2026-0770 is concrete rather than abstract, because the process runs as root in default containers and the observed in-the-wild activity takes AWS credentials and cloud metadata, so the blast radius is the cloud identity and instance-metadata reach the container holds. The control's patch-SLA-tier element cannot be exercised here and should not be recorded as though it were: the packet registers no fixed version, so the tier this asset class sits in is a compensating-control review clock, not a patch clock. Precondition: an inventory assembled from managed software distribution will miss the instances that matter, because self-hosted LLM builders are typically stood up by an application or data-science team outside it — the enumeration has to be done from the network side and reconciled, and it must be repeated, since a new instance stood up after the sweep is unprotected by definition. Distinguishing test: scan every environment for services answering on 7860 and on Langflow's API routes, reconcile the result against the asset register, and confirm each hit carries a recorded exposure decision — an ISMS asset register that lists no Langflow at all makes the technical-vulnerability-management attestation vacuous rather than clean.",
+        "evidence": "Packet attack_vector states the unauthenticated POST reaches exec() and gives remote code execution as the Langflow process, 'root in default containers', and that observed exploitation steals AWS credentials and cloud metadata. live_patch_notes name the compensating controls as blocking/removing /api/v1/validate/code exposure, network-isolating Langflow, and not exposing port 7860 to untrusted networks, and record that no patched version exists (GHSA-g22f-v6f7-2hrh, first_patched_version: null) with the CISA KEV required action being vendor mitigation or product discontinuation. CWE-829; poc_available true; active_exploitation confirmed; CISA KEV-listed 2026-07-21.",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-network-security",
+          "UK-CAF-B4"
+        ]
+      }
+    ]
   },
   "CVE-2021-27137": {
     "name": "DD-WRT Stack-Based Buffer Overflow Vulnerability",
@@ -53071,7 +53536,34 @@
         "adequate": false,
         "gap": "Technical-vulnerability management depends on an asset inventory that enumerates the firmware build; embedded UPnP services on edge routers are routinely outside that inventory, so the fixed changeset is never applied."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-030",
+        "name": "PERIMETER-DEVICE-ZERODAY-SLA-TIER",
+        "description": "A DD-WRT router is the WAN/LAN trust boundary, and this CVE reaches it with a single unauthenticated UDP datagram: an M-SEARCH to SSDP/1900 whose oversized ST header is strcpy'd unbounded into a fixed 128-byte stack buffer in ssdp.c's ssdp_msearch, giving code execution with no credential, no session and no user interaction in the path. A generic operating-system patch window is the wrong instrument for it twice over — the device is usually not an OS line item on the estate at all, and the remediation the packet records is flashing a DD-WRT build at or after change set 45724, which needs a device reboot and, on most fleets, a hands-on flash per unit that no monthly cycle can absorb. So the tier requirement has to be met by the isolation half while the flash schedule runs: disable UPnP and block UDP/1900 at the WAN edge, as the packet's interim mitigation states. Precondition, stated because this control is routinely recorded as though it closed the surface: the WAN-edge block stops only the internet-side attacker. UPnP exists to serve the client network, so on a unit where the service stays enabled the vulnerable handler remains reachable from every host on the LAN, including an already-compromised IoT device — the filter bounds the attacker population, it does not remove the path. Where UPnP is not operationally needed, disabling it removes the listener and is the stronger action; where a LAN service depends on UPnP port mapping, neither lever applies and reaching change set 45724 is the only remedy for that unit. Distinguishing test: from the WAN side and again from a client on the LAN, send an M-SEARCH whose ST value exceeds 128 bytes and confirm no SSDP handler answers either — a patch attestation covering servers and workstations reports green while the boundary device itself still parses attacker-controlled datagrams.",
+        "evidence": "Packet vector: the DD-WRT UPnP/SSDP handler (ssdp.c, ssdp_msearch) performs an unbounded strcpy of a user-supplied oversized ST-header value into a fixed 128-byte stack buffer, allowing an unauthenticated attacker to overflow the buffer and achieve code execution; attack_vector confirms the trigger is an unauthenticated UDP M-SEARCH to SSDP/1900. CWE-121, CVSS 8.1, RWEP 70, poc_available true, active_exploitation confirmed, CISA KEV-listed 2026-07-21. live_patch_available false, with live_patch_notes stating there is no live-patch mechanism for router firmware, that remediation is flashing a DD-WRT build at or after change set 45724 which requires a device reboot, and that the interim mitigation is to disable UPnP and block UDP/1900 at the WAN edge.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIST-800-53-SI-2",
+          "ISO-27001-2022-A.8.8",
+          "UK-CAF-B4",
+          "NIS2-Art21-network-security"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-032",
+        "name": "PERIMETER-COMPROMISE-REBUILD-NOT-PATCH",
+        "description": "The packet does not leave post-exploitation hypothetical: exploitation is confirmed and it names both the weaponizer and its objective — the Gafgyt/C0XMO botnet, weaponizing this overflow in 2026 to recruit routers as DDoS bots. On any DD-WRT unit whose SSDP/1900 was reachable during the exposure window, flashing change set 45724 is a patch and not a remediation, because the attacker already held code execution on the device. Flashing router firmware normally preserves the stored configuration, so an attacker-set administrative credential, DNS server, port forward or startup script carries across the upgrade untouched and the operator finishes with a device that reports the fixed build and is still attacker-held. The runbook for this device must therefore default to capturing the running configuration for analysis, reflashing to the fixed build with the stored configuration erased rather than migrated, re-entering settings from a known-good record, and rotating the administrative credential and any wireless or upstream credential the unit held — patch-in-place is the failure mode, not the plan. Detection, keyed to what an exploit doing exactly what this packet describes actually emits: inbound SSDP M-SEARCH datagrams to UDP/1900 whose ST header runs past the 128-byte buffer, and sustained outbound flood traffic sourced from the router itself, which is what recruitment as a DDoS bot produces. Precondition: both signals need the router's WAN-side traffic visible to something upstream, and a branch or home-office unit with no upstream telemetry produces no such record — there the safe default is to treat any unit that was exposed as compromised and rebuild it rather than to read the absence of an alert as evidence it was not hit.",
+        "evidence": "Packet attack_vector: an unauthenticated UDP M-SEARCH to SSDP/1900 with an oversized ST header overflows a 128-byte stack buffer in DD-WRT's UPnP handler via unbounded strcpy, enabling code execution; 'weaponized in 2026 by the Gafgyt/C0XMO botnet to recruit routers as DDoS bots'. active_exploitation confirmed; CISA KEV-listed 2026-07-21; poc_available true; CWE-121; RWEP 70. live_patch_notes record no live-patch mechanism for router firmware and that remediation is flashing a DD-WRT build at or after change set 45724, requiring a device reboot.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIST-800-53-SI-2",
+          "ISO-27001-2022-A.8.8",
+          "UK-CAF-B4"
+        ]
+      }
+    ]
   },
   "CVE-2025-68686": {
     "name": "Fortinet FortiOS Exposure of Sensitive Information to an Unauthorized Actor Vulnerability",
@@ -54527,7 +55019,33 @@
         "adequate": false,
         "gap": "A.8.7 protection against malware assumes reputation/warning controls raise friction on downloaded payloads, but this bypass removed the SmartScreen prompt entirely, letting Magniber-style loaders execute on first click until the fix was deployed."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-120",
+        "name": "DOWNLOAD-PROVENANCE-ENFORCEMENT",
+        "description": "Exploitation of this CVE needs the victim to open the file, and the whole trick sits in the delivery: the payload is wrapped so the Mark-of-the-Web is stripped or evaded, and with no untrusted-origin mark on the file SmartScreen's Open File - Security Warning never appears, so the download executes on the first open with no reputation check in the path. On an estate that has not yet taken the July 2023 Windows cumulative security update and restarted, the enforceable control is upstream of the handler: the mail gateway, the web proxy and the untrusted file-share boundary apply the untrusted-origin tag themselves, rather than leaving the marking to whatever container the attacker chose, and that tag must survive extraction from an archive, mounting from an ISO or VHD, and renaming, so the wrapper cannot launder a downloaded payload into looking locally originated. Precondition, and it has two halves that both have to be said. First, boundary-applied provenance reaches only the ingress paths those boundaries actually control: a payload written to disk by a cloud-sync client, arriving on removable media, or delivered over any channel that never applies the tag lands unmarked regardless of policy, and on that host the cumulative update and its reboot is the only thing that closes the path. Second, this addresses the strip variant — where the crafted wrapper causes the marked file to be mishandled rather than unmarked, the tag is present and the check still fails, and again only the update closes it. Distinguishing test: deliver the payload through each ingress path nested inside an archive, extract it on a managed workstation, and confirm the extracted file still carries the untrusted-origin mark and still raises the warning — a user-application-hardening attestation covering macro and add-in settings says nothing about whether provenance survived the container.",
+        "evidence": "Packet attack_vector: 'A crafted file wrapped to strip or evade the Mark-of-the-Web bypasses SmartScreen's Open File - Security Warning, so a downloaded payload executes without the reputation prompt once the user opens it.' Vector and name identify it as a Microsoft Windows Defender SmartScreen security-feature-bypass vulnerability; CWE-693; CVSS 8.8; RWEP 65; poc_available true; active_exploitation confirmed; CISA KEV-listed 2023-07-11. patch_available true with live_patch_available false, and live_patch_notes stating there is no vendor live-patch mechanism and that remediation is the July 2023 Windows cumulative security update, which requires a reboot to take effect.",
+        "gap_closes": [
+          "AU-Essential-8-App-Hardening",
+          "NIST-800-53-SI-3",
+          "ISO-27001-2022-A.8.7",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-041",
+        "name": "PROTECTION-MECHANISM-FAILURE-CLASS-REGRESSION",
+        "description": "What failed here is a defense rather than a parser: CWE-693, a SmartScreen security-feature bypass, where the file is wrapped so the Mark-of-the-Web is stripped or evaded and the Open File - Security Warning simply does not appear. That changes what counts as proof of remediation — for a memory-safety bug the build number is reasonable evidence, but for a bypassed protection mechanism the only evidence the July 2023 Windows cumulative security update fixed anything is that the warning now fires for the wrapping that previously suppressed it. Applied to this CVE, the MotW/SmartScreen class needs a standing battery in the detonation chamber that replays the wrapper forms which strip or evade the mark, executed on every cumulative-update deployment rather than once against this CVE, and held alongside the EDR and AppLocker/WDAC rules that have to carry the load whenever the prompt does not fire — because a reputation prompt the user can be walked past was never an execution control to begin with. Precondition: the battery proves only what it replays, so a wrapper form nobody added to it is untested no matter how green the run looks; and because this update takes effect only after a reboot, a host that installed it and has not restarted will pass a patch-inventory query while still executing the payload silently. Distinguishing test: after the update lands and the host has restarted, open the wrapped payload on a managed workstation and confirm the Open File - Security Warning is raised — a patch register showing the July 2023 cumulative update deployed is a record of an install, not evidence that the bypass class is closed.",
+        "evidence": "Packet name and vector: Microsoft Windows Defender SmartScreen Security Feature Bypass Vulnerability / 'Windows SmartScreen Security Feature Bypass Vulnerability', CWE-693. attack_vector: a crafted file wrapped to strip or evade the Mark-of-the-Web bypasses SmartScreen's Open File - Security Warning, so a downloaded payload executes without the reputation prompt once the user opens it. active_exploitation confirmed; CISA KEV-listed 2023-07-11; poc_available true; CVSS 8.8; RWEP 65. live_patch_available false and live_patch_notes record no vendor live-patch mechanism, with remediation being the July 2023 Windows cumulative security update, which requires a reboot to take effect.",
+        "gap_closes": [
+          "NIS2-Art21-patch-management",
+          "NIST-800-53-SI-3",
+          "ISO-27001-2022-A.8.7",
+          "UK-CAF-B4"
+        ]
+      }
+    ]
   },
   "CVE-2023-35311": {
     "name": "Microsoft Outlook Security Feature Bypass Vulnerability",
@@ -54588,7 +55106,31 @@
         "adequate": false,
         "gap": "A.8.8 technical-vulnerability management would prioritize CVE-2023-35311 on its KEV listing, but until the Office update is applied the compensating control (user heeding the security prompt) is exactly what the flaw neutralizes."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-041",
+        "name": "PROTECTION-MECHANISM-FAILURE-CLASS-REGRESSION",
+        "description": "The thing this CVE defeats is itself a protective control: a time-of-check-time-of-use race suppresses the Outlook Security Notice, the prompt that normally precedes activation of risky content in a message. Bound to Outlook, the control means that prompt is treated as a protection-mechanism class with its own regression battery — every known primitive that has suppressed or bypassed the Security Notice is replayed against a staged Outlook build on each Office and Windows update deployment, and the outcome is recorded as a pass/fail on the mechanism itself rather than inferred from the build number. That distinction is the whole point here: the compliance record for this entry is a patch level, while the thing an operator actually depends on is whether the warning still appears, and nothing in a flaw-remediation attestation exercises the prompt. The same applies downward — any phishing or user-awareness control in the estate that assumes the user is warned before activating message content must be recorded as not operative on clients below the fix, because the packet's exploitation path is precisely that the warning is not shown. Distinguishing test: on a staged client at or above the July 2023 update, deliver the crafted-message primitive and confirm the Security Notice is displayed before any risky content can be activated; recording the update as installed is not a demonstration that the prompt returned. Precondition, and it is a real limit: a regression battery can only replay bypass primitives that are already known and reproducible, so it will pass against a novel suppression of the same prompt. It also verifies rather than remediates — it gives an operator nothing during the window before the July 2023 Microsoft security update is applied and Outlook is restarted.",
+        "evidence": "The packet's attack vector is a time-of-check-time-of-use race (cwe_refs: CWE-367) in Outlook that lets a maliciously crafted message bypass the Outlook Security Notice prompt, 'so the protective warning that normally precedes activation of risky content is not shown, easing execution of a phished payload' — the bypassed item is a protection mechanism, which is what puts this entry in that class. The entry is CISA KEV-listed 2023-07-11 with active_exploitation 'confirmed' (rwep_score 47, cvss 8.8); poc_available is false, so prioritization rests on the KEV listing and confirmed exploitation rather than on a public exploit. patch_available is true and live_patch_available is false, with live_patch_notes stating the fix ships in the July 2023 Microsoft security update and applies on the next Office/Windows update cycle, typically requiring an application restart or reboot.",
+        "gap_closes": [
+          "NIST-800-53-SI-2",
+          "ISO-27001-2022-A.8.8",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-001",
+        "name": "CISA-KEV-RESPONSE-SLA",
+        "description": "For this entry the KEV clock opened 2023-07-11 and the packet records no live-patch path, so the only remediation is the July 2023 Microsoft security update — which the packet says applies on the next Office/Windows update cycle and typically requires an application restart or reboot. Bound to Outlook, that makes the completion definition the load-bearing half: the clock runs to a restarted Outlook client on every mailbox, not to an update marked approved, downloaded or installed in the management console. A workstation that took the July 2023 update while Outlook stayed running continuously still carries the vulnerable code, and a long-uptime desktop with a never-closed mail client is exactly where that deferral hides from a patch report. Standard 14- and 30-day patch SLAs miss this twice: the window is wrong for a KEV-listed flaw under confirmed exploitation, and the completion measure is wrong for a fix that only takes effect on client restart. Distinguishing test: report, per mailbox client, whether the Outlook process has been restarted since the update was applied, not merely which Office build is present on disk. Precondition: this control is a clock and a completion definition, not a mitigation. It does not reduce exposure during the window, and the packet names no vendor mitigation and no compensating control for this entry — until the update lands and Outlook restarts, nothing operator-side restores the suppressed prompt, so any process that treats the Security Notice as a barrier should be assumed to have none.",
+        "evidence": "cisa_kev is true with kev_date 2023-07-11 and active_exploitation 'confirmed' (rwep_score 47, cvss 8.8). patch_available is true; live_patch_available is false; live_patch_notes: 'No live-patch mechanism; the fix ships in the July 2023 Microsoft security update and applies on the next Office/Windows update cycle, typically requiring an application restart or reboot.' The packet records no vendor mitigation or compensating control for this entry. The gaps cited against it are patch- and vulnerability-management controls (ASD Essential Eight 'Patch operating systems', ISO/IEC 27001:2022 A.8.8, NIST SP 800-53 SI-2 Flaw Remediation, NIS2 Art. 21 vulnerability handling and disclosure).",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIS2-Art21-patch-management",
+          "NIST-800-53-SI-2"
+        ]
+      }
+    ]
   },
   "CVE-2023-36874": {
     "name": "Microsoft Windows Error Reporting Service Privilege Escalation Vulnerability",
@@ -54649,7 +55191,31 @@
         "adequate": false,
         "gap": "A.8.8 technical-vulnerability management may downgrade a 7.8 local-only LPE versus network CVEs, yet as a live-exploited escalation link in intrusion chains it warranted emergency patching that a CVSS-vector-only triage would delay."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-145",
+        "name": "ACTIVELY-EXPLOITED-OS-LPE-REMEDIATION-WINDOW-ENFORCEMENT",
+        "description": "The packet puts the attacker at ordinary standard-user privilege on the host: they plant a malicious wermgr.exe and use NTFS junctions or hard links in a WER-controlled path so the Windows Error Reporting service, which trusts the link target, launches that binary as SYSTEM. For this CVE the control means the July 2023 Windows cumulative update is driven across every affected host on the clock that opened with the 2023-07-11 KEV listing rather than folded into the next monthly rollup, with completion measured per host as installed build at or above the fixed build AND the reboot actually taken — the packet records no live-patch path and a fix applied through the standard servicing stack that requires a reboot, so a host that staged the update and has not restarted still runs the vulnerable code and must be counted as exposed. Enumerate first the hosts where non-administrative users hold interactive local sessions — shared workstations, session hosts, jump boxes, build and developer machines — because on those the exploit's only stated precondition, an ordinary local user, is the normal operating state rather than an anomaly. The control's second half is the load-bearing one here, and it is exactly why the least-privilege gap is recorded against this entry: the attacker holds no privilege they were not legitimately granted, so removing local-administrator rights or tightening per-account scoping does not contain the escalation, and an AC-6 attestation passes cleanly while the flaw stays fully exploitable. Precondition on the interim lever: restricting which accounts may log on interactively bounds the population that can attempt this, but it does not close the path — any account that legitimately reaches the desktop satisfies the precondition in full, and on a shared session host that is every user. And because active exploitation is confirmed and a public PoC exists, a host that carried untrusted local users during the exposure window needs forensic triage and credential rotation rather than closure on the patch: the update repairs the link handling but removes nothing an attacker already installed while holding SYSTEM.",
+        "evidence": "The packet's attack vector: 'A standard user plants a malicious wermgr.exe and uses NTFS junctions/hard links in a WER-controlled path so the Windows Error Reporting service, which trusts the link target (CWE-59), launches the attacker binary as NT AUTHORITY SYSTEM.' cwe_refs is CWE-59. cisa_kev true, kev_date 2023-07-11, active_exploitation 'confirmed', poc_available true, rwep_score 68, cvss 7.8. patch_available is true and live_patch_available is false; live_patch_notes: 'No vendor live-patch mechanism; remediation requires the July 2023 Windows cumulative update, which is applied via the standard servicing stack and requires a reboot.' NIST SP 800-53 AC-6 (Least Privilege) is among the framework gaps citing this entry.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIS2-Art21-patch-management",
+          "NIST-800-53-AC-6",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-018",
+        "name": "SCANNER-PAPER-COMPLIANCE-TEST",
+        "description": "A scan that reports this host 'patched' because the July 2023 cumulative update appears in its servicing record is paper compliance twice over. First, the packet states the update is applied through the standard servicing stack and requires a reboot, so a host with the update staged and no restart reports patched while still running the vulnerable Windows Error Reporting path — reboot-pending state has to be reported alongside installed build for every host in the remediation population, or the dashboard converts an exposed host into a compliant one. Second, and more fundamental for this CVE, the exploitable condition is a filesystem-link behaviour rather than a version: a standard user creating an NTFS junction or hard link in a WER-controlled path so the service launches a planted binary from a location that user can write. No version query observes that. The operational test, keyed to the behaviour the packet documents: on a staging host at or above the fixed build and rebooted, run as a standard user, plant a binary, redirect a WER-controlled path at it through a junction or hard link, and confirm the Error Reporting service does not launch the linked target with SYSTEM privileges. Precondition: this is a verification method, not a mitigation. It tells the operator whether remediation actually landed on the build they tested and nothing about hosts on a different build, it requires a staging host on which the plant can be attempted safely, and it says nothing about a host already compromised before the test was run — confirmed exploitation means the estate can hold hosts where the answer is 'the primitive worked, weeks ago'.",
+        "evidence": "live_patch_notes record the fix as 'the July 2023 Windows cumulative update, which is applied via the standard servicing stack and requires a reboot', with live_patch_available false — so an installed-but-unrebooted host is still vulnerable. The attack vector describes the exploitable condition as a standard user planting a malicious wermgr.exe and using NTFS junctions/hard links in a WER-controlled path that the Windows Error Reporting service trusts (CWE-59) to launch the binary as SYSTEM, which is a behaviour rather than a version state. active_exploitation is 'confirmed' and poc_available is true (kev_date 2023-07-11). The entry's cited gaps include ISO/IEC 27001:2022 A.8.8 (management of technical vulnerabilities) and ASD Essential Eight patching.",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "AU-Essential-8-Patch"
+        ]
+      }
+    ]
   },
   "CVE-2022-31199": {
     "name": "Netwrix Auditor Insecure Object Deserialization Vulnerability",
@@ -55248,7 +55814,31 @@
         "adequate": false,
         "gap": "A.8.8 technical-vulnerability management struggles with mobile-baseband/DSP components an organisation cannot directly patch; the ELF-into-DSP capability was an OEM-only fix, leaving a coverage gap until the SMR propagated to each device."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-126",
+        "name": "MOBILE-OS-MINIMUM-PATCH-LEVEL-ENFORCEMENT",
+        "description": "The packet's exploitation path is local: a process already running on the handset abuses the Samsung DSP driver's missing ELF validation to load arbitrary libraries into the DSP, gaining code execution in DSP context as a step in an on-device privilege-escalation chain. Remediation is the SMR Mar-2021 Release 1 or later firmware update, which reboots the device. Bound to this estate, the control means the fixed security-patch level is enforced as an access condition rather than published as a statistic: a managed Samsung handset below SMR Mar-2021 Release 1 is denied mail, VPN and document access by policy until it reaches that level, so the exposure is removed from the organization's data instead of being recorded against it. Enrolment has to read the device's actual security-patch level, because that — not the Android version and not the model — is what distinguishes a handset still carrying the vulnerable DSP driver. For any handset that cannot be brought to SMR Mar-2021 Release 1 or later, reaching a fixed build is not available at all: the terminal state is replacement and removal from the estate on a dated schedule, and a requirement that ends at 'every device reports the fixed level' would mark such a handset compliant only by leaving it out of the count. Distinguishing test: enrol a handset pinned below SMR Mar-2021 Release 1 and confirm policy actually denies it access to protected resources — an estate that surfaces the stale patch level on a report while the device keeps its mail and VPN sessions has recorded the exposure, not removed it. Precondition, and it is the limit that matters here: this is an access condition on organizational data, and the packet's exploitation path presupposes a local process already running on the device. It therefore neither removes that process nor reverses code already loaded into the DSP on a handset exploited before enrolment — a device suspected of that belongs on the incident path, not the access-policy path — and it is a holding measure only for the window before the firmware update and its reboot land.",
+        "evidence": "The packet's vector: 'A vulnerability in DSP driver prior to SMR Mar-2021 Release 1 allows attackers load arbitrary ELF libraries inside DSP', and its attack vector: 'A local process abuses the Samsung DSP driver's missing ELF validation to load arbitrary libraries into the DSP, gaining code execution in the DSP context as a stepping stone in an on-device privilege-escalation chain.' cwe_refs is CWE-912. cisa_kev true with kev_date 2023-06-29 and active_exploitation 'confirmed'; poc_available false; rwep_score 48, cvss 6.7. patch_available is true and live_patch_available is false; live_patch_notes: 'No live-patch mechanism for the Samsung DSP driver; remediation is applying the SMR Mar-2021 Release 1 (or later) firmware update, which reboots the device.' The packet states no end-of-support status for the affected handsets, so whether a given unit can reach the fixed level is a per-device question the inventory has to answer.",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-056",
+        "name": "MOBILE-ENDPOINT-MDM-ENFORCED-KEV-SLA",
+        "description": "The KEV clock on this entry opened 2023-06-29 under confirmed exploitation, and the packet records no live-patch mechanism for the Samsung DSP driver — the only remediation is the SMR Mar-2021 Release 1 or later firmware update, which reboots the device. The control means that update is driven from the MDM/EMM managing the handset on the KEV clock with user deferral disallowed, and completion measured per device as the security-patch level the handset itself reports after it restarts. The restart is the step this particular remediation loses: a firmware update that reboots a phone someone carries is the update a user postpones indefinitely, and a handset that has downloaded it without restarting still runs the vulnerable driver while appearing actioned in the console. Priority should follow the packet rather than the 6.7 CVSS band — the packet describes DSP-context code execution as a stepping stone in an on-device privilege-escalation chain, so this is a chain link on a device holding organizational mail, tokens and credentials rather than a standalone low-severity item, and confirmed in-the-wild exploitation is what sets the clock, not the score. Precondition: enforcement reaches only handsets the MDM/EMM actually enrols and can drive to a firmware level. An unenrolled or personally-managed device is outside this control entirely and has to be handled by conditional access instead, and a handset for which SMR Mar-2021 Release 1 or later is not obtainable cannot be remediated by this control at all — that unit belongs on the replacement path, since there is no build the operator can drive it to.",
+        "evidence": "cisa_kev true, kev_date 2023-06-29, active_exploitation 'confirmed' (rwep_score 48 against cvss 6.7). live_patch_available is false with live_patch_notes: 'No live-patch mechanism for the Samsung DSP driver; remediation is applying the SMR Mar-2021 Release 1 (or later) firmware update, which reboots the device.' The attack vector describes the DSP-context code execution as 'a stepping stone in an on-device privilege-escalation chain'. The gaps citing this entry are patch- and vulnerability-management controls (ASD Essential Eight 'Patch operating systems', NIST SP 800-53 SI-2 Flaw Remediation, NIS2 Art. 21 vulnerability handling).",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIS2-Art21-vulnerability-management",
+          "NIST-800-53-SI-2"
+        ]
+      }
+    ]
   },
   "CVE-2021-25372": {
     "name": "Samsung Mobile Devices Improper Boundary Check Vulnerability",
@@ -56292,7 +56882,44 @@
         "adequate": false,
         "gap": "A.8.8 technical-vulnerability management for network appliances often lacks asset visibility into edge firewalls, so the unauthenticated RCE persisted on unpatched Zyxel devices well past the vendor's 2023-05-24 fix while botnet enrollment proceeded."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-030",
+        "name": "PERIMETER-DEVICE-ZERODAY-SLA-TIER",
+        "description": "The affected units are the trust boundary itself — Zyxel ATP, USG FLEX, USG FLEX 50(W), USG20(W)-VPN, VPN series and ZyWALL/USG firewalls — and the packet places the overflow in the ID-processing function reachable by an unauthenticated attacker, so the device that terminates the perimeter is the device that runs the attacker's code. For these units the control means the fixed firmware branch (above 5.36 Patch 1, or above 4.73 Patch 1 on the ZyWALL/USG line) is driven on the KEV clock that opened 2023-06-05 rather than into the next appliance-maintenance window, and completion is measured per unit by the firmware version actually running after the reboot the flash requires. The packet records no live-patch path, so a unit with the image staged but not yet rebooted is still running the vulnerable ID-processing code and must be counted exposed. Where the reboot cannot be taken immediately, the control's alternative is isolation of the vulnerable interface, which the packet identifies as restricting WAN-side management and VPN access. Precondition, and this is where that interim mitigation is routinely over-claimed: restricting WAN-side reachability bounds who can send the crafted request, but on a USG20(W)-VPN or VPN-series unit whose function is terminating remote-access VPN from the internet, the VPN service must keep answering untrusted sources, so for that interface the restriction removes nothing and only the firmware flash does. A unit in that state is on the accelerated clock, not on a compensating control.",
+        "evidence": "Packet: CWE-120 buffer overflow in the ID-processing function, reachable by an unauthenticated attacker for DoS or remote code execution. Affected branches ATP 4.32–5.36 Patch 1, USG FLEX 4.50–5.36 Patch 1, USG FLEX 50(W) 4.25–5.36 Patch 1, USG20(W)-VPN 4.25–5.36 Patch 1, VPN 4.30–5.36 Patch 1, ZyWALL/USG 4.25–4.73 Patch 1. CISA KEV 2023-06-05, active_exploitation confirmed, CVSS 9.8, RWEP 57. patch_available true; live_patch_available false, with live_patch_notes stating remediation requires flashing fixed firmware above the affected 5.36 Patch 1 / 4.73 Patch 1 branches, which reboots the device, and naming restriction of WAN-side management/VPN access as the recommended interim mitigation.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-patch-management",
+          "NIST-800-53-SI-2",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-032",
+        "name": "PERIMETER-COMPROMISE-REBUILD-NOT-PATCH",
+        "description": "The packet gives unauthenticated code execution on the firewall with confirmed in-the-wild exploitation from the 2023-06-05 KEV listing, so for any unit that was internet-reachable during the exposure window the open question is not whether the vulnerable function is still present but whether the device is still the operator's. Flashing the fixed firmware replaces the vulnerable ID-processing code; it does not remove configuration an attacker wrote, accounts an attacker added, or credentials an attacker read out of the device — and on an ATP / USG FLEX / USG20(W)-VPN / VPN-series / ZyWALL unit that means VPN pre-shared keys, remote-access user credentials and administrator passwords. For this CVE the control means any such unit that answered from the WAN during the window is handled as suspected-compromised: its configuration exported for comparison against a known-good baseline rather than carried forward, the device rebuilt onto the fixed firmware from a clean configuration, and every credential the device held rotated. Precondition: the rebuild is only as good as the baseline it restores from — a configuration exported after the exposure window and re-imported unchanged reinstates whatever the attacker left in it — and the reboot the flash requires is not a compromise-recovery step on its own. The packet records poc_available false; that is not a triage input here, because active exploitation is separately recorded as confirmed, so absence of a public exploit is not evidence a given unit was untouched.",
+        "evidence": "Packet: unauthenticated attacker sends a crafted request that overflows the buffer in the firewall's ID-processing function, causing a crash or executing attacker-controlled code on the device. CISA KEV 2023-06-05; active_exploitation confirmed; poc_available false; CVSS 9.8; RWEP 57. patch_available true, live_patch_available false, with the vendor fix being a firmware flash that reboots the device.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIS2-Art21-patch-management",
+          "NIST-800-53-SI-2"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-038",
+        "name": "MITIGATION-ACTIVE-PATCH-PENDING-VERDICT-CLASS",
+        "description": "This entry has two remediation states that a vulnerability register normally collapses into one. State (a) is the Zyxel firmware above the 5.36 Patch 1 / 4.73 Patch 1 branches, flashed and rebooted, which removes the overflow. State (b) is the packet's own named interim mitigation — WAN-side management and VPN access restricted — with the vulnerable firmware still running, so the ID-processing overflow is intact and reachable by anything still inside the permitted scope. For these firewalls the control means state (b) is recorded as a time-bound compensating control with a dated action item to reach the flash, never as 'patched within SLA', and the register carries the firmware version running on each unit rather than a remediation flag. The distinguishing test is per unit and keys on what the device reports rather than on what the register asserts: read back the firmware version actually running and the current WAN access rules, and confirm the recorded verdict matches. An estate whose report shows the KEV item closed while units still run 5.36 Patch 1 behind an access rule has recorded a mitigation as a fix. Precondition: this is accounting, not defence — it changes nothing on the device. Its value is that it stops the interim state ageing out of view, which on unmanaged edge units is how the exposure outlives the availability of the fix.",
+        "evidence": "Packet: patch_available true and live_patch_available false, with live_patch_notes stating that remediation requires flashing fixed firmware above the affected 5.36 Patch 1 / 4.73 Patch 1 branches (which reboots the device) and that restricting WAN-side management/VPN access is the recommended interim mitigation — two distinct states named in the same entry. CISA KEV 2023-06-05, active_exploitation confirmed, CVSS 9.8.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2"
+        ]
+      }
+    ]
   },
   "CVE-2023-34362": {
     "name": "Progress MOVEit Transfer SQL Injection Vulnerability",
@@ -56597,7 +57224,31 @@
         "adequate": false,
         "gap": "A.8.8 technical-vulnerability management struggles to enforce timely WebKit updates on personally-owned or embedded devices that rely on WebKit for HTML processing, leaving a known-exploited info-disclosure flaw unremediated past the KEV due date."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-056",
+        "name": "MOBILE-ENDPOINT-MDM-ENFORCED-KEV-SLA",
+        "description": "The fix for this WebKit read does not ship as one update — the packet names watchOS 9.5, tvOS 16.5, macOS Ventura 13.4, iOS and iPadOS 16.5, iOS and iPadOS 15.7.6, and Safari 16.5. An enforcement policy written only against iOS/iPadOS and current macOS therefore leaves the Apple TVs, the Watches, and any Mac taking Safari 16.5 as a separate update outside the SLA entirely, while the compliance view reads clean. For this CVE the control means the enforced minimum is expressed per train against the specific fixed build rather than as 'latest major version': a device on the 15.x line is remediated by 15.7.6 and must be driven there rather than written off as unsupportable, and a Mac left on an older macOS is remediated by Safari 16.5. The packet records no live-patch path and states the fix requires installing the update and restarting the device, so the completion signal is the post-restart build each device reports — an update downloaded or sitting at 'pending restart' still runs the vulnerable HTML-processing path and is not remediated. Precondition: this reaches only devices the management channel actually enrols and can compel. Hardware in the estate that no management channel holds is not covered by this control at all, and the remaining lever there is withholding access to organizational data until the device reports a fixed build — not carrying it as a standing exception.",
+        "evidence": "Packet: CWE-125 out-of-bounds read; the vector states the issue is fixed in watchOS 9.5, tvOS 16.5, macOS Ventura 13.4, iOS 15.7.6 and iPadOS 15.7.6, Safari 16.5, iOS 16.5 and iPadOS 16.5, and that processing web content may disclose sensitive information. CISA KEV 2023-05-22; active_exploitation confirmed; CVSS 6.5; RWEP 45. patch_available true; live_patch_available false, with live_patch_notes stating there is no live-patch mechanism and that the listed fixes require installing the update and restarting the device.",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-patch-management",
+          "NIST-800-53-SI-2",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-121",
+        "name": "MOBILE-ZERO-CLICK-HARDENING",
+        "description": "The packet's outcome is an information disclosure — sensitive memory including pointers usable to defeat ASLR — not code execution, and the CVSS of 6.5 reflects that. Read as a chain input rather than a standalone bug, the priority inverts: the leak is what makes a following memory-corruption step reliable, so the population that matters is the cohort the operator has already designated as high-risk. For those users the control means a standing reduced-attack-surface posture on their Apple devices, so untrusted web content is not processed automatically by WebKit, narrowing the delivery path into the out-of-bounds read during the window between the 2023-05-22 KEV listing and the completed install-and-restart of watchOS 9.5 / tvOS 16.5 / macOS Ventura 13.4 / iOS and iPadOS 16.5 or 15.7.6 / Safari 16.5. Two preconditions, both load-bearing. The posture narrows what reaches the parser; it does not remove the flaw — content the user deliberately opens still reaches WebKit's HTML processing — so this is a holding measure for the pre-update window, not a substitute for the fixed build and its restart. And it only helps if it was already on when the content arrived: assigning it after a device has rendered the attacker's page changes nothing about what was already read out of that process, which is an incident-response matter rather than a configuration one. Scope it to the designated cohort and to the Apple products the packet names; extending it estate-wide is a usability cost that adds no coverage of this CVE.",
+        "evidence": "Packet: maliciously crafted web content drives WebKit's HTML processing past a buffer boundary, reading out-of-bounds memory and disclosing sensitive information such as pointers used to defeat ASLR (CWE-125). Vector states processing web content may disclose sensitive information and that Apple is aware of a report the issue may have been actively exploited; active_exploitation recorded as confirmed. CISA KEV 2023-05-22; CVSS 6.5; RWEP 45. live_patch_available false, with the fix requiring installation of the listed builds and a device restart.",
+        "gap_closes": [
+          "AU-Essential-8-App-Hardening",
+          "UK-CAF-B4"
+        ]
+      }
+    ]
   },
   "CVE-2023-32373": {
     "name": "Apple Multiple Products WebKit Use-After-Free Vulnerability (CVE-2023-32373)",
@@ -56872,7 +57523,30 @@
         "adequate": false,
         "gap": "A.8.15 logging controls require logs not to expose sensitive data and to be access-protected; here the kernel writes raw pointers into a log a privileged process can read, so the logging control as implemented leaks the exact addresses that defeat ASLR."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-126",
+        "name": "MOBILE-OS-MINIMUM-PATCH-LEVEL-ENFORCEMENT",
+        "description": "The packet pins remediation to one named build level: kernel pointers are printed in the log file prior to SMR May-2023 Release 1, and the only remediation it records is Samsung SMR May-2023 Release 1 or later delivered by device OTA, which reboots the handset. Bound to this estate, the control means that SMR level is enforced as an access condition on every managed Samsung handset — a device reporting below SMR May-2023 Release 1 is refused organizational mail, VPN and document access until it reports at or above it — rather than appearing as a stale row on a patch-compliance report. Enforcement rather than reporting is the point here because this entry's own severity signals argue for deferral: the packet gives CVSS 4.4, RWEP 44 and poc_available false, while the flaw's function is to hand a privileged local process the kernel addresses that defeat ASLR, which is an enabling step whose value to an attacker never shows up in its own score. Distinguishing test: enrol a handset pinned below SMR May-2023 Release 1 and confirm the policy actually denies it protected resources; an estate that surfaces the stale patch level on a dashboard while the device keeps mail and VPN has recorded the exposure rather than removed it. Preconditions, all of which are ways this control is over-claimed: it reaches only enrolled, managed handsets, so an unenrolled or personally-owned device is bounded solely by withholding organizational data from it; the packet records live_patch_available false, so a handset that has downloaded the OTA but not completed it and its reboot is still running the logging kernel and must be counted as exposed, not remediated; and where the OTA carrying SMR May-2023 Release 1 or later is not offered for a given model, no enforcement action produces the fixed build at all — for those models the terminal state is withdrawal of organizational data and replacement of the handset, not an access-policy exception. Because active_exploitation is confirmed, a device suspected of having already been through the chain this disclosure primitive feeds belongs on the incident path; reaching the fixed build does not evict what a prior compromise established.",
+        "evidence": "Packet vector: \"Kernel pointers are printed in the log file prior to SMR May-2023 Release 1 allows a privileged local attacker to bypass ASLR.\" Packet live_patch_notes: \"No live-patch mechanism; remediation requires Samsung SMR May-2023 Release 1 or later via the device OTA update, which reboots the handset\" — with patch_available true and live_patch_available false. CISA KEV listed 2023-05-19 with active_exploitation confirmed, against CVSS 4.4, RWEP 44 and poc_available false, which is the deprioritization pressure the packet records. The packet's attack_vector states the addresses are used to defeat KASLR and provide the address-disclosure primitive needed for reliable follow-on kernel exploitation in a spyware chain. The entry cites NIST-800-53-SI-2, NIS2-Art21-patch-management and UK-CAF-B4 as insufficient.",
+        "gap_closes": [
+          "NIST-800-53-SI-2",
+          "NIS2-Art21-patch-management",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-059",
+        "name": "SENSITIVE-DATA-IN-LOGS-LINT",
+        "description": "On this CVE the disclosure is the log's content. The CWE is CWE-532 and the packet's vector says kernel pointers are printed in the log file, which a privileged local process then reads to learn kernel addresses. That is precisely the question a logging control does not ask: an A.8.15-style attestation verifies that logs are produced, retained, time-referenced and access-protected, and every one of those passes on a handset whose log is itself the address-disclosure primitive, because the defect is what the platform chose to write rather than who may read it. Bound to this product, the requirement is that log content be treated as an output classified by what it reveals. On the platform side that is a supplier requirement, since only the OEM build stops the pointers being written: mobile-platform acceptance and re-acceptance must ask whether kernel and diagnostic logs carry raw kernel addresses, not only whether logging is enabled and protected. On the operator side, every pipeline that collects these device logs — MDM diagnostic capture, crash and bug-report upload, log-forwarding agents — inherits those addresses off the handset, so on any device below SMR May-2023 Release 1 those captures widen the set of readers and must be scoped, access-restricted and retention-bounded as sensitive material rather than handled as routine telemetry. Distinguishing test: on a handset below SMR May-2023 Release 1, read the log through the privileged local path the packet describes and confirm whether raw kernel pointers appear; a logging attestation reporting collection coverage, retention period and log access control passes cleanly while the pointers sit in the file. Precondition: this control governs what is written and who receives it — it does not repair the kernel. Nothing operator-side prevents the pointers being logged on a device below the fixed build, so restricting log handling bounds readership without removing the primitive, and the privileged local process the packet describes is already inside that boundary by definition. It is a holding measure for the window before SMR May-2023 Release 1 and its reboot land, not a substitute for them.",
+        "evidence": "Packet cwe_refs: CWE-532 (Insertion of Sensitive Information Into Log File), matching the entry name \"Samsung Mobile Devices Insertion of Sensitive Information Into Log File Vulnerability\". Packet vector: kernel pointers are printed in the log file prior to SMR May-2023 Release 1, allowing a privileged local attacker to bypass ASLR. Packet attack_vector: \"The kernel writes raw kernel pointers into a log file; a privileged local process reads those entries to learn kernel addresses and defeat KASLR.\" The entry cites ISO-27001-2022-A.8.15 (Logging) and AU-Essential-8-App-Hardening (User application hardening) as insufficient. CISA KEV listed 2023-05-19 with active_exploitation confirmed; patch_available true with live_patch_available false.",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.15",
+          "AU-Essential-8-App-Hardening"
+        ]
+      }
+    ]
   },
   "CVE-2023-25717": {
     "name": "Multiple Ruckus Wireless Products CSRF and RCE Vulnerability",
@@ -56994,7 +57668,42 @@
         "adequate": false,
         "gap": "A.8.8 technical-vulnerability management often deprioritizes local-only CVEs, yet this one converts any low-privilege foothold into root, so treating it as low-urgency (no network vector) misjudges its post-compromise value."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-145",
+        "name": "ACTIVELY-EXPLOITED-OS-LPE-REMEDIATION-WINDOW-ENFORCEMENT",
+        "description": "polkit arbitrates privileged D-Bus operations on effectively every Linux install, and the packet's path needs nothing but an unprivileged local account: repeatedly call a privileged D-Bus method and kill the client mid-request until polkit races, treats the disconnected caller as root, authorizes the action, and lets the caller create a new local administrator. For this CVE the control means the distribution's fixed polkit package is driven across every Linux host on the clock that opened with the 2023-05-12 KEV listing, with completion measured per host against the polkit that is actually running rather than against 'deployed' in a management console. The packet makes that measurement both cheap and easy to skip: there is no kernel live-patch to consider because polkit is a userspace daemon, and the flaw closes once polkitd restarts with no full reboot — so there is no maintenance window to schedule and no user-visible outage to defer, and a host that took the package but never restarted polkitd is still running the vulnerable daemon and must be counted exposed. Enumerate first the hosts where this exploit's precondition is the normal operating state rather than an anomaly: multi-user shell servers, jump hosts, shared build machines and CI runners, where every interactive account already satisfies 'unprivileged local user'. The least-privilege gap cited on this entry cannot contain the escalation — the attacker is a legitimately authorized user, the privilege transition happens inside polkit's credential check, and an attestation that every account holds only the rights it needs passes cleanly while the flaw stays fully exploitable. Because active exploitation is confirmed and a PoC is public, remediation also has to look backwards: the outcome the packet names is a new local administrator account, and that account survives the package update untouched. Reconcile local accounts and administrative group membership on hosts that were exposed instead of closing the finding on the package version.",
+        "evidence": "Packet: CWE-863 / CWE-754, CISA KEV 2023-05-12, active_exploitation confirmed, poc_available true, CVSS 7.8, RWEP 64. attack_vector: 'An unprivileged local user repeatedly calls a privileged D-Bus method and kills the client mid-request; polkit races and treats the disconnected caller as root, authorizing the action and allowing creation of a new administrator account.' Vector: the flaw 'could be used by an unprivileged local attacker to, for example, create a new local administrator.' patch_available true; live_patch_available false with live_patch_notes 'No kernel live-patch applies — polkit is a userspace daemon; remediation is the distribution's fixed polkit package, after which restarting polkitd (no full reboot) closes the flaw.' NIST-800-53-AC-6 (Least Privilege) is recorded among the citing gaps.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-vulnerability-management",
+          "NIST-800-53-AC-6"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-146",
+        "name": "USERSPACE-PRIVILEGED-DBUS-RACE-DETECTION",
+        "description": "polkit is a userspace daemon, so what this control means for this CVE is host audit or eBPF rules on the privilege-escalation behaviour rather than kernel-exploit indicators — and the packet describes that behaviour precisely enough to key on it. The exploit is a race, and a race is repetitive: an unprivileged local user calls a privileged D-Bus method over and over and kills the client mid-request until the timing lands. The rule therefore keys on that shape — bursts of short-lived invocations of privileged D-Bus methods from a non-root uid that terminate before the call completes — paired with the privilege transition the packet names as the result: a new local account appearing with administrative rights, seen as writes to the local account and group files or as account-management utilities running outside a change window. Both halves are needed. Repeated D-Bus calls alone can be benign; an account creation alone arrives with no context, and it is the pairing inside a short window that distinguishes this exploit from either. Note what will not see it: nothing is compiled, no module is loaded, no setuid binary changes, no process crashes, and the work is done by an ordinary user session calling a system daemon through its normal interface — so file-integrity monitoring and signature-based endpoint tooling have no artifact to match, and alerting on crashes or on named exploit tooling would miss an attempt that behaves exactly as the packet describes. Precondition: this requires host audit or eBPF telemetry already collected and shipped off-host before the attempt, which on the multi-user and shared hosts most exposed to this flaw is precisely where it is least often enabled; a rule authored after the fact against telemetry nobody was collecting produces nothing. And detection does not prevent the escalation and does not remove an administrator account already created — it bounds the window to alert-and-response time during the period before the fixed polkit package and the polkitd restart land.",
+        "evidence": "Packet attack_vector: 'An unprivileged local user repeatedly calls a privileged D-Bus method and kills the client mid-request; polkit races and treats the disconnected caller as root, authorizing the action and allowing creation of a new administrator account.' Vector: 'polkit could be tricked into bypassing the credential checks for D-Bus requests, elevating the privileges of the requestor to the root user'; the stated example outcome is creating 'a new local administrator'. active_exploitation confirmed, poc_available true, CISA KEV 2023-05-12. live_patch_notes states polkit is a userspace daemon and no kernel live-patch applies.",
+        "gap_closes": [
+          "UK-CAF-B4",
+          "NIST-800-53-AC-6"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-018",
+        "name": "SCANNER-PAPER-COMPLIANCE-TEST",
+        "description": "For this CVE the paper-compliance failure has one exact shape: a scanner reads the installed polkit package version, finds it at or above the distribution's fixed build, and reports the host remediated — while the polkitd process that actually arbitrates D-Bus authorization is still the pre-update binary, because the packet states the flaw closes only after polkitd restarts. A package-version query cannot tell those two hosts apart, and on a long-uptime multi-user server — the same population where any interactive account already satisfies this exploit's 'unprivileged local user' precondition — they diverge as a matter of routine, since installing the package does not by itself take the daemon down. The operational test is therefore not the package query but the running daemon: for every host reported patched, show that the polkitd currently running was started after the fixed package was installed, or verify the running daemon's on-disk image directly. Anything that stops at the package version is an attestation about the filesystem, not about the process making the authorization decision. Precondition: this test verifies remediation, it does not perform it — a host that fails it is remediated by restarting polkitd, which the packet records as needing no full reboot — and it says nothing about whether the host was already exploited. With exploitation confirmed in the wild and the documented outcome being a new local administrator account that the package update does not remove, that is a separate question answered by reconciling local accounts, not by any version check.",
+        "evidence": "Packet live_patch_notes: 'No kernel live-patch applies — polkit is a userspace daemon; remediation is the distribution's fixed polkit package, after which restarting polkitd (no full reboot) closes the flaw.' patch_available true, live_patch_available false. active_exploitation confirmed, poc_available true, CISA KEV 2023-05-12. attack_vector places the attacker as an unprivileged local user; the vector's stated outcome is creation of a new local administrator.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-vulnerability-management"
+        ]
+      }
+    ]
   },
   "CVE-2014-0196": {
     "name": "Linux Kernel Race Condition Vulnerability",
@@ -57055,7 +57764,30 @@
         "adequate": false,
         "gap": "A.8.8 technical-vulnerability management that scores by CVSS treated this 5.5 issue as low urgency, understating that it yields reliable local root; the low base score let it linger unpatched despite a public exploit and KEV listing."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-018",
+        "name": "SCANNER-PAPER-COMPLIANCE-TEST",
+        "description": "On this CVE the distance between \"patched\" and \"not vulnerable\" is exactly one reboot, and that is where the compliance record goes wrong. The packet's remediation is upgrading to a fixed Linux kernel past 3.14.3 and rebooting, with live_patch_available false. A scanner that reads the installed kernel package version marks a host compliant the moment the fixed package lands, while the kernel actually executing — the one still carrying the vulnerable n_tty_write path in the LECHO and !OPOST case — is unchanged until the machine restarts. The requirement for this entry is that the vulnerability verdict be taken from the running kernel, comparing the live release against the fixed build, and that \"fixed package installed, reboot pending\" be reported as its own exposed state with a dated restart rather than as a pass. The entry's compliance risk is the compound of two ordinary program behaviours: the packet gives CVSS 5.5 against RWEP 71 with poc_available true and confirmed exploitation, so a program that prioritises by base score defers the kernel update, and a program that then measures by package version closes the deferral without the reboot ever happening. Enumerate the population where the deferral matters most first — hosts where unprivileged users hold interactive shell sessions, since the exploit needs a local account and a pty, which on a shared or multi-user host is the normal operating state rather than an anomaly. Distinguishing test: take a host that has installed the fixed kernel package but has not restarted, run the scan, and confirm it reports that host as still exposed; a scan returning \"patched\" there is measuring the package rather than the kernel in memory, and the fleet's remediation count for this CVE is inflated by every host in that state. Precondition: this control corrects what is measured and what action is scheduled — it removes nothing on its own, and a host it correctly reports as exposed stays exploitable until the restart is taken.",
+        "evidence": "Packet live_patch_notes: \"Standard remediation is upgrading to a fixed Linux kernel (past 3.14.3) and rebooting,\" with patch_available true and live_patch_available false; the same note records that no specific vendor live-patch for this CVE is confirmed. Packet vector: the n_tty_write function in drivers/tty/n_tty.c in the Linux kernel through 3.14.3 does not properly manage tty driver access in the \"LECHO & !OPOST\" case. Packet scoring: CVSS 5.5 against rwep_score 71, with poc_available true, CISA KEV listing 2023-05-12 and active_exploitation confirmed. Packet attack_vector: a local user races writes to a pty to escalate to root. The entry cites NIST-800-53-SI-2, AU-Essential-8-Patch, ISO-27001-2022-A.8.8 and NIS2-Art21-patch-management as insufficient.",
+        "gap_closes": [
+          "NIST-800-53-SI-2",
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-patch-management"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-003",
+        "name": "KERNEL-EXPLOITATION-DETECTION",
+        "description": "This is the control for the window the reboot deferral leaves open, and for this CVE it has to key on the two behaviours the packet actually documents. First, the race itself: an unprivileged process opens a pty, puts the line discipline into the echo-on with output-processing-off configuration the packet names, and then drives sustained concurrent long writes to it from more than one thread. An interactive session does not produce that — a human-driven terminal does not emit multi-threaded bulk writes to its own pty — so the write pattern is the signal that fires while the exploit is looping, which matters because a race is attempted many times before it wins. Second, the transition it is racing toward: a euid change to 0 in a process descended from an unprivileged login session with no preceding setuid execution or authorization event to account for it. Both halves are required — the write-storm rule sees the attempts, the uid-transition rule sees a success — and the alert has to reach an operator inside 60 seconds because the response to a win is isolating the host, not queueing a ticket. Do not build the rule around the crash. The packet does record memory corruption and system crash as the alternative outcome, but the run that wins the race produces no crash at all, so a rule keyed on kernel oops volume alone is blind to exactly the case that matters and would read as clean during a successful escalation. Distinguishing test: the packet records poc_available true, so run the public exploit against a staging host on an unfixed kernel and confirm the write-storm rule fires during the attempt loop and the uid-transition rule fires on success — a rule set validated only against a synthetic uid change or a forced panic has not been shown to see this exploit. Precondition: detection does not restore the containment this flaw breaks; it tells you the boundary failed. It requires a low-privilege foothold to be visible to the sensor in the first place, and it is a compensating measure for hosts awaiting the restart, not a reason to extend the deferral.",
+        "evidence": "Packet attack_vector: \"Two threads race concurrent long writes to a pty in the n_tty_write LECHO/!OPOST path, overflowing the tty buffer to corrupt kernel memory and escalate a local user to root (or crash the machine).\" Packet vector: the flaw allows local users to cause a denial of service (memory corruption and system crash) or gain privileges by triggering a race condition involving read and write operations with long strings; cwe_refs CWE-362. Packet poc_available true supports a runnable validation. Packet live_patch_available false with remediation requiring a reboot establishes the deferral window this covers. CISA KEV listed 2023-05-12, active_exploitation confirmed, rwep_score 71. The entry cites UK-CAF-B4 (System security) as insufficient, and the lesson's framework_coverage records the B4 gap as the least-privilege containment assumption failing on shared and multi-tenant Linux hosts that skipped the kernel bump.",
+        "gap_closes": [
+          "UK-CAF-B4"
+        ]
+      }
+    ]
   },
   "CVE-2010-3904": {
     "name": "Linux Kernel Improper Input Validation Vulnerability (CVE-2010-3904)",
@@ -57116,7 +57848,41 @@
         "adequate": false,
         "gap": "A.8.9 configuration management should enforce a hardened kernel-module baseline, but default configurations that leave the rarely-used RDS module loadable let a local user reach the unvalidated rds_page_copy_user copy, so configuration drift/default-on modules sustained the root-escalation exposure."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-009",
+        "name": "KERNEL-MODULE-INVENTORY-AND-DISABLE",
+        "description": "The packet names this mitigation itself: the RDS module can be blacklisted or unloaded on hosts that do not use it, removing the reachable code path without an immediate kernel update. Bound to this CVE, the control means resolving, per host, which of three states net/rds is actually in — already loaded, loadable on demand, or built into the running kernel image — and blacklisting it everywhere RDS carries no business function, because the exploit's first move is an unprivileged user opening an AF_RDS socket, and on a host where rds is merely auto-loadable that socket call is what pulls the unvalidated rds_page_copy_user path into the kernel in the first place. The inventory is the load-bearing half, because \"we blacklisted rds\" is a fleet-level assertion and the three states behave nothing alike. Preconditions, which is exactly where this control gets over-claimed as removing the surface outright: a blacklist prevents the on-demand load and does nothing on a host where rds is already loaded — that host also needs the module unloaded, and the unload fails while anything holds an RDS socket open, so a host that cannot be quiesced enough to unload is a reboot rather than a blacklist. A blacklist does nothing at all where RDS is built into the kernel image rather than shipped as a module: there is no module to refuse, the AF_RDS path exists from boot, and the only remediation there is the fixed kernel — address validation in rds_page_copy_user, in 2.6.36 or a distro backport — with the reboot the packet says it requires. And hosts that genuinely use RDS cannot take this mitigation in any form; they are patch-and-reboot items with no interim compensating control from this direction. Distinguishing test: on a representative host, as an unprivileged user, attempt to open an AF_RDS socket and confirm it fails, and confirm rds is absent from the loaded module set — not merely that a blacklist file is present in configuration. A configuration-management attestation showing the blacklist deployed passes on a host that loaded rds before the file landed and on a host whose kernel has RDS built in, and on both the arbitrary-kernel-write path is still fully reachable by any local account.",
+        "evidence": "Packet live_patch_notes: \"the kernel fix (address validation in rds_page_copy_user, in 2.6.36 or a distro backport) requires a reboot to take effect. As a non-reboot mitigation, the RDS module can be blacklisted/unloaded on hosts that do not use it, removing the reachable code path without an immediate kernel update,\" with patch_available true and live_patch_available false. Packet vector: rds_page_copy_user in net/rds/page.c in the RDS implementation in the Linux kernel before 2.6.36 does not properly validate addresses obtained from user space, allowing local users to gain privileges via crafted use of sendmsg and recvmsg. Packet attack_vector: \"A local user opens an AF_RDS socket and issues crafted sendmsg/recvmsg calls,\" giving an arbitrary kernel write leveraged to escalate to root. poc_available true; CISA KEV listed 2023-05-12; active_exploitation confirmed; rwep_score 70. The entry cites NIST-800-53-CM-7, ISO-27001-2022-A.8.9, AU-Essential-8-App-Hardening and UK-CAF-B4 as insufficient; the lesson's framework_coverage records that stock distributions left the module auto-loadable and that only patching to 2.6.36+ or disabling the module closes it.",
+        "gap_closes": [
+          "NIST-800-53-CM-7",
+          "ISO-27001-2022-A.8.9",
+          "AU-Essential-8-App-Hardening",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-038",
+        "name": "MITIGATION-ACTIVE-PATCH-PENDING-VERDICT-CLASS",
+        "description": "This entry produces the middle state the control exists to name, and produces it by design rather than by accident. The packet records no vendor live patch, a kernel fix that requires a reboot, and a module blacklist or unload as the interim mitigation — so a host with rds blacklisted is not patched: the unvalidated rds_page_copy_user copy is still present in the running kernel and becomes reachable again the instant anything loads the module. The requirement is that compliance records for this CVE carry that as its own verdict — mitigation active, vulnerable kernel running, reboot pending, with a dated action item for the fixed kernel — and never merge it into the same remediated bucket as a host actually running 2.6.36 or a backported kernel. The distinction is operational rather than clerical, because the events that silently return a mitigated host to full exposure are all routine maintenance: a kernel package upgrade or image rebuild that reinstates a default module configuration, a host restored from a baseline predating the blacklist, or a new host provisioned from an older image. None of those generate a vulnerability event, because from the scanner's point of view nothing about the host changed, so a fleet can drift back toward exposure while its remediation count for this CVE holds steady. Distinguishing test: take any host recorded as remediated on CVE-2010-3904 and check whether its record distinguishes \"module blacklisted\" from \"running a fixed kernel\"; if the report cannot tell those apart, the remediation count includes hosts that are one configuration-management run from exploitable by any local account. Precondition: this control changes what is recorded and what action is scheduled — it removes no code path itself, and a host sitting in the mitigated state carries the residual risk for as long as the reboot is deferred, which on a long-lived server is the entire point of the deferral.",
+        "evidence": "Packet live_patch_notes records both states explicitly: no vendor-provided live patch is published for this CVE, the kernel fix requires a reboot to take effect, and the RDS module can be blacklisted/unloaded as a non-reboot mitigation on hosts that do not use it. patch_available true with live_patch_available false. Packet attack_vector: a local user opens an AF_RDS socket and issues crafted sendmsg/recvmsg calls for an arbitrary kernel write leveraged to root; poc_available true and active_exploitation confirmed, CISA KEV listed 2023-05-12. The entry cites NIS2-Art21-vulnerability-management and ISO-27001-2022-A.8.9 as insufficient; the lesson's framework_coverage records the A.8.9 gap as configuration drift and default-on modules sustaining the exposure.",
+        "gap_closes": [
+          "NIS2-Art21-vulnerability-management",
+          "ISO-27001-2022-A.8.9"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-017",
+        "name": "BUG-FAMILY-MITIGATION-PERSISTENCE",
+        "description": "The blacklist and the patch are not equivalent removals here, which is why dropping the blacklist once the fixed kernel lands is a net loss of posture on hosts that never used RDS. The packet's fix is address validation in rds_page_copy_user — one function, one missing check. Refusing to load net/rds removes reachability of the whole subsystem, of which the unvalidated copy is one defect. So on any host where the inventory established that RDS carries no business function, the blacklist should be treated as a permanent least-functionality decision that survives the kernel upgrade, not as a temporary measure to be reverted when the patched kernel boots. The concrete failure this prevents is a rollback framed as cleanup: an operator reversing the modprobe configuration as part of closing the CVE ticket, restoring an unprivileged-reachable protocol implementation to a host that had no use for it, and doing so at exactly the moment the change looks safest. Distinguishing test: after the fixed kernel is deployed and the host has restarted, confirm the blacklist entry is still in place and that an unprivileged AF_RDS socket open still fails — a remediation workflow that reverts compensating configuration on patch confirmation will show a host that is patched for this CVE and once again carrying the full net/rds surface. Precondition: this governs retention of a control that was available in the first place. It gives nothing to a host where RDS is in genuine use or is built into the kernel image, since no blacklist existed there to retain, and it is not a reason to defer the fixed kernel — the retention applies after the patch lands, not instead of it.",
+        "evidence": "Packet live_patch_notes names both the fix and the mitigation: \"the kernel fix (address validation in rds_page_copy_user, in 2.6.36 or a distro backport) requires a reboot to take effect. As a non-reboot mitigation, the RDS module can be blacklisted/unloaded on hosts that do not use it, removing the reachable code path.\" Packet vector scopes the defect to rds_page_copy_user in net/rds/page.c failing to validate addresses obtained from user space. Packet attack_vector: an unprivileged local user reaches it by opening an AF_RDS socket. The entry cites NIST-800-53-CM-7 (Least Functionality) and ISO-27001-2022-A.8.9 (Configuration management) as insufficient; the lesson's framework_coverage records the CM-7 gap as stock distributions leaving the module auto-loadable so hosts that never used RDS still exposed the arbitrary-kernel-write path.",
+        "gap_closes": [
+          "NIST-800-53-CM-7",
+          "ISO-27001-2022-A.8.9"
+        ]
+      }
+    ]
   },
   "CVE-2015-5317": {
     "name": "Jenkins User Interface (UI) Information Disclosure Vulnerability",
@@ -57177,7 +57943,30 @@
         "adequate": false,
         "gap": "A.8.8 technical-vulnerability management may rank a 2015 info-disclosure as low priority, yet its KEV status means the leak of otherwise-restricted job/build names remains a live reconnaissance vector on any Jenkins below 1.638 / LTS 1.625.2."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-129",
+        "name": "ENTERPRISE-ERP-MANAGEMENT-FUNCTION-AUTHENTICATION",
+        "description": "Jenkins enforces per-item access control at the view layer, and this CVE is one view not enforcing it: the packet has a direct request to the Fingerprints pages returning the names of jobs and builds the requester is not authorized to see. Bound to this product, the control means every Jenkins page and endpoint resolves the caller's permission on the specific item it is about to name — the fingerprint record, the job, the build — before it renders, rather than inheriting the verdict that allowed the caller to reach Jenkins at all; and the Jenkins web surface is segmented so callers with no operational need cannot issue that direct request in the first place. This is exactly why the access-enforcement gap recorded on this entry passes its attestation while the path stays open: the deployment's authorization matrix can be configured precisely as intended, reviewed, and evidenced, and this page still never consults it, so the finding is invisible to any audit that inspects the permission configuration rather than the response body. Precondition: the per-page authorization behaviour is what upgrading to 1.638 or later (LTS 1.625.2 or later) establishes — this control states the property to verify, it does not implement it — and the packet records no live-patch mechanism, so until that release and the Jenkins service restart land, restricting who can reach the controller (removing anonymous read, fronting it with authenticated access, taking it off any internet-reachable path) bounds the population that can issue the request and leaves it fully effective for every developer, contractor and service account that legitimately reaches Jenkins, which on a shared controller is most of the organization. Distinguishing test: with an account authorized to read exactly one job, request the Fingerprints pages on a staging controller and confirm no job or build name outside that account's authorization appears in the response — an attestation that Jenkins role-based authorization is configured and reviewed passes cleanly while this page keeps answering.",
+        "evidence": "Packet: CWE-200. Vector: 'The Fingerprints pages in Jenkins before 1.638 and LTS before 1.625.2 might allow remote attackers to obtain sensitive job and build name information via a direct request.' attack_vector: 'A direct request to the Jenkins Fingerprints pages returns names of jobs and builds the user is not authorized to see, disclosing CI/CD pipeline structure for reconnaissance.' NIST-800-53-AC-3 (Access Enforcement) is recorded among the citing gaps. patch_available true; live_patch_available false with live_patch_notes requiring the upgrade to 1.638 or later (LTS 1.625.2 or later) and a Jenkins service restart.",
+        "gap_closes": [
+          "NIST-800-53-AC-3",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-001",
+        "name": "CISA-KEV-RESPONSE-SLA",
+        "description": "The remediation this packet records is narrow and fully specified: upgrade Jenkins to 1.638 or later (LTS 1.625.2 or later) and restart the Jenkins service. There is no live-patch mechanism, so a controller whose WAR was replaced but whose service was never restarted is still serving the vulnerable pages and has to be counted exposed rather than patched. What puts this on the KEV clock is not the score — RWEP 36, CVSS 7.5, and the packet records no public PoC — but that exploitation is confirmed in the wild while the technique the packet describes is a direct request to a page, so the absence of published exploit code is no barrier to its use and no reason to let the item sit in a low-severity queue. The population this control has to reach is precisely the one a severity-ordered backlog deprioritizes: a controller still on a pre-1.638 line at the 2023-05-12 KEV listing is not a host waiting on a vendor, it is a host nobody owns — a team-run build server, an instance left behind after a migration to another CI system, a controller inside a project VM. So the first obligation here is enumeration: every Jenkins controller reachable on the estate, with its reported version and its service start time, discovered from the network rather than taken from the CI team's inventory. Precondition: the compensating-control branch of this SLA is weaker than usual for this CVE. The disclosure happens on an ordinary page request, so there is no distinctive request pattern for a proxy or WAF rule to key on, and restricting reachability only narrows who can ask — it does not stop the page answering. Remediation is also not restorative: job and build names disclosed before the upgrade stay disclosed, and the packet identifies the value of that disclosure as reconnaissance on CI/CD pipeline structure, so a controller with confirmed exposure warrants review of what that structure reveals rather than closure on the version bump alone.",
+        "evidence": "Packet: CISA KEV 2023-05-12, active_exploitation confirmed, poc_available false, CVSS 7.5, RWEP 36, CWE-200. patch_available true, live_patch_available false, live_patch_notes 'No live-patch mechanism; remediation requires upgrading Jenkins to 1.638 or later (LTS 1.625.2 or later) and restarting the Jenkins service.' Vector names the affected versions ('before 1.638 and LTS before 1.625.2') and the technique ('via a direct request'); attack_vector states the disclosure is of 'CI/CD pipeline structure for reconnaissance'.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-vulnerability-management"
+        ]
+      }
+    ]
   },
   "CVE-2016-3427": {
     "name": "Oracle Java SE and JRockit Unspecified Vulnerability",
@@ -57482,7 +58271,39 @@
         "adequate": false,
         "gap": "A.8.8 technical-vulnerability management relies on an accurate component inventory (SBOM); without one, organizations could not enumerate where the vulnerable Log4j library ran, leaving a ransomware-associated RCE unremediated well past the KEV due date."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-042",
+        "name": "INCOMPLETE-PATCH-SEQUENCE-SEVERITY-MULTIPLIER",
+        "description": "This CVE exists because a prior remediation was incomplete: the packet's vector states the fix that addressed CVE-2021-44228 in Log4j 2.15.0 was incomplete in certain non-default configurations, leaving the same JNDI-lookup primitive reachable through attacker-controlled Thread Context Map data. That makes the operator's problem a records problem as much as a code problem — the version move to 2.15.0 is already closed on the flaw-remediation ledger, so the follow-on step reads as a routine minor bump rather than as an unremediated actively-exploited flaw. Applied to this CVE, an application on 2.15.0 must be scored as unremediated for this primitive and driven to 2.16.0 (Java 8) or 2.12.2 (Java 7) with the affected Java application restarted, on the clock that opened with the 2023-05-01 KEV listing. The sequence position also sets the standard for what counts as a fix: the packet credits the fixed releases with removing support for message lookup patterns and disabling JNDI functionality by default — a capability removal — so a narrower fix on this same primitive should be treated as provisionally incomplete until the capability-removing version is in service. Distinguishing test: pull the remediation evidence for this CVE and check what it points at; if it points at the December 2021 Log4Shell response rather than at a per-application version reading of 2.16.0 or 2.12.2 on the runtime classpath, the attestation is recording the first CVE's closure as this one's.",
+        "evidence": "Packet vector: the fix to address CVE-2021-44228 in Apache Log4j 2.15.0 was incomplete in certain non-default configurations, allowing attackers with control over Thread Context Map (MDC) input data to craft a JNDI Lookup pattern; the packet states Log4j 2.16.0 (Java 8) and 2.12.2 (Java 7) fix the issue by removing support for message lookup patterns and disabling JNDI functionality by default. live_patch_notes: remediation requires upgrading Log4j to 2.16.0 (Java 8) or 2.12.2 (Java 7) and restarting the affected Java application. CISA KEV-listed 2023-05-01, active_exploitation confirmed, PoC available, CVSS 9.0, RWEP 74, patch_available true, live_patch_available false.",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-038",
+        "name": "MITIGATION-ACTIVE-PATCH-PENDING-VERDICT-CLASS",
+        "description": "The packet itself separates the two states this control requires an audit to distinguish: removing the JndiLookup class from the classpath is recorded as a stopgap mitigation, not a vendor live-patch, while remediation is upgrading to 2.16.0 (Java 8) or 2.12.2 (Java 7) and restarting the affected Java application. Applied to this estate, every Java application carrying Log4j needs an explicit verdict of (a) fixed version resolved on the runtime classpath and the application restarted, (b) still on an affected version with the JndiLookup class stripped, or (c) neither — and (b) has to appear as a time-bound compensating-control state with a dated action item, never as a patched-per-SLA outcome. Preconditions on (b), which are the reason it cannot be recorded as closure: the stopgap holds only while that class stays absent from the artifact, so a redeploy, a dependency re-resolution, a rollback to an earlier build, or a restore that reinstates the original library returns the application to full exposure silently, and the removal is applied per artifact rather than centrally, so it has to be re-verified on every build rather than once. It also leaves the affected version in place, so the capability changes the packet credits the fixed releases with — removal of message lookup patterns and JNDI disabled by default — are not in effect for anything else that reaches the same code. And because the fix takes effect only on restart, an application whose dependency was upgraded but whose process has not been restarted is still state (c) dressed as state (a). Distinguishing test: for each application recorded in state (b), rebuild it from its own pipeline and inspect the produced artifact for the JndiLookup class; if it comes back, the mitigation was a one-off edit to a deployed artifact and the next deployment reopens the flaw.",
+        "evidence": "Packet live_patch_notes: 'No live-patch mechanism; remediation requires upgrading Log4j to 2.16.0 (Java 8) or 2.12.2 (Java 7) and restarting the affected Java application. The stopgap of removing the JndiLookup class from the classpath is a mitigation, not a vendor live-patch.' patch_available true, live_patch_available false. Packet vector states the fixed releases remove support for message lookup patterns and disable JNDI functionality by default. CISA KEV-listed 2023-05-01, active_exploitation confirmed, PoC available, CVSS 9.0, RWEP 74.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "NIST-800-53-SI-2"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-021",
+        "name": "TIER-3-DEPENDENCY-INVENTORY",
+        "description": "Log4j is not an installed product on this estate but a library reached through Java applications — the packet's remediation is expressed per application (upgrade Log4j and restart the affected Java application), so how much of the estate is remediated is exactly how much of it the dependency inventory can see. Applied to this CVE, the inventory has to record per application: the Log4j version actually resolved onto the runtime classpath rather than only the version named as a direct dependency; the Java track, because the packet gives different fixed releases for Java 8 (2.16.0) and Java 7 (2.12.2), so a single target version applied estate-wide leaves one track wrong; and whether the logging configuration uses a non-default PatternLayout with a Context Lookup or a Thread Context Map pattern (%X, %mdc, %MDC), which the packet gives as the condition under which attacker-controlled MDC data reaches the JNDI lookup. That last item is a configuration property rather than a code property, so an application marked not-exploitable on configuration grounds is re-exposed the moment someone adds a correlation-id pattern to a log format, with no dependency change to trigger a re-scan — it has to be re-checked on logging-configuration change, not only on dependency change. Distinguishing test: take applications whose declared direct dependencies name no Log4j at all and enumerate their resolved runtime classpath; an inventory built from declared direct dependencies reports full coverage while transitively-included copies sit on an affected version.",
+        "evidence": "Packet live_patch_notes: remediation requires upgrading Log4j to 2.16.0 (Java 8) or 2.12.2 (Java 7) and restarting the affected Java application — two fixed releases on two Java tracks, applied per application. Packet vector conditions exploitation on a non-default Pattern Layout with either a Context Lookup (for example, $${ctx:loginId}) or a Thread Context Map pattern (%X, %mdc, or %MDC) plus attacker control over MDC input data. The entry cites NIS2 Art. 21 supply-chain security measures as a framework gap. CISA KEV-listed 2023-05-01, active_exploitation confirmed, PoC available, CVSS 9.0, RWEP 74.",
+        "gap_closes": [
+          "NIS2-Art21-supply-chain",
+          "ISO-27001-2022-A.8.8"
+        ]
+      }
+    ]
   },
   "CVE-2023-21839": {
     "name": "Oracle WebLogic Server Unspecified Vulnerability (CVE-2023-21839)",
@@ -57726,7 +58547,30 @@
         "adequate": false,
         "gap": "A.8.8 technical vulnerability management could only react — the sandbox escape was a zero-day, so remediation depended on rapidly deploying Chrome 112.0.5615.137 across all managed endpoints."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-057",
+        "name": "BROWSER-MANAGED-UPDATE-NO-DEFERRAL",
+        "description": "The packet gives a fixed build — Chrome/Chromium 112.0.5615.137 — and a KEV-listed sandbox escape under confirmed exploitation, which is the case a staged enterprise update ring is worst at: the ring exists to soak a release across a validation window, and the window is the exposure. For this CVE the control means the security-channel update to 112.0.5615.137 or later is pushed to every managed Chrome/Chromium install on the clock that opened with the 2023-04-21 KEV listing, with user deferral disabled, and completion measured by the build each browser instance is actually running rather than by the version approved or staged in the management console; the packet records no live-patch mechanism, so an instance still executing a pre-112.0.5615.137 build carries the vulnerable Skia even once the newer build is on disk. This is also the control that carries the exposure the user-application-hardening gap cited on this entry cannot: hardening settings govern what content the renderer is allowed to process, whereas the packet places this defect at the step after that — reached from a renderer the attacker already controls — so no hardening setting narrows the Skia integer overflow, and a hardening attestation passes cleanly while the escape stays available. Priority follows the packet rather than the absence of a public exploit: no PoC is recorded, but exploitation is confirmed, so waiting for a public PoC to raise the priority inverts the evidence. Precondition: this reaches only installs the browser update channel manages; a Chromium build shipped inside another product does not take the Chrome update and is not covered here — it belongs to the embedded-component inventory and parity requirement instead.",
+        "evidence": "Packet vector: 'Integer overflow in Skia in Google Chrome prior to 112.0.5615.137 allowed a remote attacker who had compromised the renderer process to potentially perform a sandbox escape via a crafted HTML page.' attack_vector: from a renderer process the attacker already controls, the crafted page drives the overflow to break out of the sandbox and execute in the higher-privileged browser process. live_patch_notes: no live-patch mechanism; remediation requires updating Chrome/Chromium to 112.0.5615.137 or later. CISA KEV-listed 2023-04-21, active_exploitation confirmed, poc_available false, patch_available true, live_patch_available false, CVSS 9.6, RWEP 45. The entry cites ASD Essential Eight user application hardening as a framework gap.",
+        "gap_closes": [
+          "AU-Essential-8-App-Hardening",
+          "NIS2-Art21-patch-management",
+          "NIST-800-53-SI-2"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-144",
+        "name": "EMBEDDED-MEDIA-PARSER-LIBRARY-INVENTORY-AND-PATCH-PARITY",
+        "description": "The packet puts the defect in Skia — an embedded rendering component, not a separately installed product — and states remediation as two distinct actions: update Chrome/Chromium to 112.0.5615.137 or later, and rebuild the downstream products that embed the affected Skia, which it names as ChromeOS, Android and Flutter. Those are separate update tracks, so an attestation built on the browser fleet closes only the first action while shipped copies of the same vulnerable code stay in service. Applied here, the control means enumerating, alongside managed Chrome installs, the packet-named downstream products actually in service — ChromeOS devices, Android builds, and applications built with Flutter — and confirming each is on a build produced after the Skia fix, since each carries its own copy that the Chrome update never touches. Scope it to those products: the packet ties the affected Skia to Chrome/Chromium and to ChromeOS, Android and Flutter and provides no mapping into other software embedding Skia, so instructing operators to treat every graphics-rendering binary in the estate as an instance of this CVE manufactures rebuild and removal work against products no evidence here implicates; widen only where a verified source names another product carrying the affected component. Distinguishing test: after the Chrome fleet reports 112.0.5615.137, take one in-service Flutter-built application and one Android or ChromeOS image and show each was produced after the Skia fix — an estate whose flaw-remediation evidence is the browser inventory alone reads clean while shipped software still embeds the vulnerable code. Precondition: rebuilding is available only for products the operator builds, or for which a fixed build can be obtained; a downstream product whose supplier ships no rebuilt version is not remediated by this control and stays on the exposure register rather than being closed by it.",
+        "evidence": "Packet live_patch_notes: 'No live-patch mechanism; remediation requires updating Chrome/Chromium to 112.0.5615.137 or later and rebuilding downstream products (ChromeOS, Android, Flutter) that embed the affected Skia.' Packet vector places the integer overflow in Skia in Google Chrome prior to 112.0.5615.137, reachable via a crafted HTML page from a compromised renderer process to escape the sandbox. CISA KEV-listed 2023-04-21, active_exploitation confirmed, poc_available false, patch_available true, live_patch_available false, CVSS 9.6, RWEP 45.",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2"
+        ]
+      }
+    ]
   },
   "CVE-2017-6742": {
     "name": "Cisco IOS and IOS XE Software SNMP Remote Code Execution Vulnerability",
@@ -57787,7 +58631,30 @@
         "adequate": false,
         "gap": "A.8.8 technical-vulnerability management frequently omits network-device firmware from the vulnerability program; organizations without an inventory of IOS/IOS XE SNMP exposure could not identify or remediate the vulnerable routers before state-sponsored exploitation."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-032",
+        "name": "PERIMETER-COMPROMISE-REBUILD-NOT-PATCH",
+        "description": "Bound to a Cisco IOS / IOS XE device that was SNMP-reachable from untrusted space while this flaw was unpatched, this control needs re-scoping rather than restating, because its usual premise does not hold here: the packet's implant is in-memory (Jaguar Tooth) and the packet's remediation is a fixed Cisco software release that reboots the device, so the upgrade itself clears memory-resident code. What the upgrade does not touch is the exploit's own precondition. The packet requires the attacker to already know the device's SNMP read-only community string (v1/v2c) or its SNMPv3 user credentials, and an image upgrade rotates neither — so the same access that produced the first implant works against the rebooted device, and the second run costs the attacker nothing. It also does not surface anything the attacker changed while resident. Applied to this device, remediation is three steps, not one: before the reload discards volatile state, capture and diff the running and startup configuration against a known-good baseline (SNMP community and user definitions, ACLs, any added local accounts, any configuration that would restore attacker access after reboot); rotate every SNMP community string and SNMPv3 credential the device carries and every device that shares them, since a community string is a per-device shared secret rather than a per-operator one; then bring the device up on the fixed release from the baseline configuration rather than from the configuration the compromised device saved for itself. Distinguishing test: after the upgrade, show that the device's SNMP credentials differ from the pre-incident values and that its running configuration matches the baseline — an estate that records the fixed IOS release against the asset and closes the ticket has remediated the overflow and preserved the credential. Preconditions and limits: the configuration diff assumes a known-good baseline exists to diff against; where none does, the step degrades to an engineer reviewing the configuration line by line and is weaker for it. And rebuilding the device does not tell you what the resident implant read or reached from that position, so a device confirmed compromised still owes downstream credential rotation for anything it held or could authenticate to.",
+        "evidence": "Packet: CWE-119 buffer overflow in the SNMP implementation of Cisco IOS and IOS XE; the attacker 'must know the SNMP read only community string (SNMP version 2c or earlier) or the user credentials (SNMPv3)'; attack vector records that APT28 used it to inject the in-memory Jaguar Tooth implant; CISA KEV-listed 2023-04-19 with active_exploitation confirmed and poc_available true; RWEP 73 / CVSS 8.8; patch_available true and live_patch_available false, with the live-patch note stating remediation requires upgrading to a Cisco fixed software release per advisory cisco-sa-20170629-snmp, which reboots the device. AU-Essential-8-Patch and ISO 27001:2022 A.8.8 are recorded as insufficient for this entry — both are satisfied the moment the fixed release is installed.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-128",
+        "name": "APPSERVER-REMOTING-PROTOCOL-RESTRICTION",
+        "description": "The SNMP agent on IOS / IOS XE is the listener class this control governs, on a router rather than an application server: a non-HTTP management protocol that answers on the device itself, whose parser is the vulnerable code, and which the packet says is exploitable only by traffic directed at the affected system. The requirement for these devices is that the SNMP agent answers only the management hosts that legitimately poll it — enforced with an SNMP access list and view on the device plus infrastructure ACLs on the paths that reach it, rather than assumed from 'the management VLAN is internal' — and that the affected MIBs and OIDs the packet's mitigation names are removed from the served view wherever the estate does not poll them, which is the least-functionality step the CM-7 gap is about. Distinguishing test: from a general user VLAN and from external transit, send SNMP requests at a staging device and confirm they are dropped before the agent parses them; a device that passes a configuration audit for 'SNMPv3 with authentication enabled' while its agent still answers packets sourced from anywhere has satisfied the audit and left the overflow reachable. Preconditions, and this is where the control is easy to over-claim: restricting reachability bounds who can send the crafted packet, it does not repair the parser, and the packet states the flaw affects all versions of SNMP (1, 2c and 3) — so moving the estate to SNMPv3 with strong credentials raises the cost of obtaining a valid credential but leaves the overflow fully reachable by anyone holding one, and a leaked v1/v2c community string is a single shared secret rather than a per-operator account. The monitoring platform that must poll the device is by definition inside the permitted set, so its compromise satisfies the exploit's stated access precondition in full. This is the holding measure for the window before the fixed software release lands, and taking that release reboots the device.",
+        "evidence": "Packet: 'The vulnerability is due to a buffer overflow in the affected code area. The vulnerability affects all versions of SNMP (versions 1, 2c, and 3)'; 'Only traffic directed to the affected system can be used to exploit this vulnerability'; the live-patch note records Cisco's recommended mitigations as restricting SNMP access to trusted management hosts, disabling affected MIBs/OIDs, and enforcing SNMPv3 with strong credentials, with the fixed software release (which reboots the device) as the remediation; live_patch_available false; KEV 2023-04-19, active_exploitation confirmed, poc_available true. NIST 800-53 CM-7 (Least Functionality), NIS2 Art.21 security of network and information systems, and UK CAF B4 (System security) are the framework controls recorded as insufficient for this entry.",
+        "gap_closes": [
+          "NIST-800-53-CM-7",
+          "NIS2-Art21-network-security",
+          "UK-CAF-B4"
+        ]
+      }
+    ]
   },
   "CVE-2019-8526": {
     "name": "Apple macOS Use-After-Free Vulnerability",
@@ -57970,7 +58837,29 @@
         "adequate": false,
         "gap": "A.8.8 technical vulnerability management presumes remediation can be scheduled, but the availability of the fix for this Framework LPE is gated by Android security-patch-level delivery per OEM, so vulnerability-management SLAs cannot guarantee closure of the privilege-escalation path on affected fleets."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-126",
+        "name": "MOBILE-OS-MINIMUM-PATCH-LEVEL-ENFORCEMENT",
+        "description": "The packet's precondition is an app already on the device, and the outcome is local privilege escalation with no additional execution privileges needed and no user interaction — so on a handset below the 2023-03-01 Android security patch level there is no prompt to decline, no permission grant to review and no user decision in the path, and the only lever left is what organizational data that handset is permitted to hold. Applied to this CVE, the 2023-03-01 or later patch level must operate as an access condition across the affected Android 11, 12, 12L and 13 population — mail, VPN and document access denied to a device below it — rather than as a reported field on a patch-compliance dashboard. Distinguishing test: enrol an Android device pinned below the 2023-03-01 patch level and confirm the policy actually refuses it access to protected resources; an estate that surfaces the stale patch level on a report while the device keeps its mail and VPN access has recorded the exposure rather than removed it. Precondition, and this is where the control gets over-claimed: restricting installs to a managed catalogue and blocking sideloading raises the bar for getting the attacker's app onto the device, but it evicts nothing already installed and does not cover an app that arrived through the normal store channel, so a device suspected of already running such an app belongs on the incident path rather than the install-policy path. And because the packet frames remediation as the device receiving the 2023-03-01 or later patch level as a system update, a device model that is never offered that patch level cannot be brought into compliance by any policy the operator holds: for those handsets the access condition is a standing posture, not a bridge, and replacement is the terminal state — leaving them enrolled with an open exception marks them compliant while they stay exposed to everything found since.",
+        "evidence": "Packet vector: 'In WorkSource, there is a possible parcel mismatch. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation. Product: Android Versions: Android-11 Android-12 Android-12L Android-13 Android ID: A-220302519'. attack_vector: a malicious app already on the device triggers the Parcel serialization mismatch in WorkSource, causing the system to grant the app elevated privileges without extra permissions or user interaction. live_patch_notes: no live-patch mechanism for the Android Framework; remediation requires the device receiving the 2023-03-01 (or later) Android security patch level, installed as a system update that reboots the device. CISA KEV-listed 2023-04-13, active_exploitation confirmed, poc_available false, patch_available true, live_patch_available false, CVSS 7.8, RWEP 49.",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-056",
+        "name": "MOBILE-ENDPOINT-MDM-ENFORCED-KEV-SLA",
+        "description": "For this CVE remediation is a single event per handset — the device takes the 2023-03-01 or later Android security patch level as a system update and reboots — so the control is about driving that event across the enrolled Android 11, 12, 12L and 13 population on the clock that opened with the 2023-04-13 KEV listing, with user deferral prohibited and the restart forced rather than left to the user's convenience. Completion is counted per device from the security patch level reported after the restart, because the packet records no live-patch mechanism for the Android Framework: a handset that has downloaded and staged the update but not rebooted is still running the affected framework, and on a personal-feeling device the reboot is precisely the step users postpone, which is the specific way this remediation is recorded as done while remaining undone. Precondition, stated because this is where a mobile patch SLA is routinely over-claimed: the management policy can prohibit deferral and force installation of an update the device has already been offered, but it cannot produce a patch level that the device's OEM or carrier update channel has not shipped for that model. The SLA therefore has to be measured against per-model availability, and models with no 2023-03-01 or later build available fail this control by definition — they must be surfaced as unremediated devices carrying a replacement decision, not counted as compliant on the grounds that the policy was applied to them.",
+        "evidence": "Packet live_patch_notes: 'No live-patch mechanism for the Android Framework; remediation requires the device receiving the 2023-03-01 (or later) Android security patch level, which is installed as a system update and reboots the device.' Packet vector names the affected versions as Android-11, Android-12, Android-12L and Android-13 and states the escalation needs no additional execution privileges and no user interaction. CISA KEV-listed 2023-04-13, active_exploitation confirmed, poc_available false, patch_available true, live_patch_available false, CVSS 7.8, RWEP 49.",
+        "gap_closes": [
+          "NIST-800-53-SI-2",
+          "NIS2-Art21-patch-management"
+        ]
+      }
+    ]
   },
   "CVE-2023-29492": {
     "name": "Novi Survey Insecure Deserialization Vulnerability",
@@ -58519,7 +59408,32 @@
         "adequate": false,
         "gap": "A.8.8 would score this 'low' on CVSS alone and defer it, but the KEV listing and spyware chain show the leak's real weight as an exploitation primitive; assessing it purely by base severity mis-prioritises the mobile-fleet patch it actually needs."
       }
-    }
+    },
+    "new_control_requirements": [
+      {
+        "id": "NEW-CTRL-126",
+        "name": "MOBILE-OS-MINIMUM-PATCH-LEVEL-ENFORCEMENT",
+        "description": "The packet's exploit precondition is a non-privileged application performing valid Mali GPU operations — nothing malformed, nothing that fails — and the packet records the fix reaching devices only as the Arm-fixed Mali driver bundled into an OEM/Android security update that reboots the handset. Applied to this driver, the control means each enrolled handset's security-patch level is evaluated as an access condition on what the device may reach — mail, VPN, document and MDM-brokered data withheld from any device below the level that carried the fixed driver — rather than surfaced as a red row on a patch-compliance report while the device keeps its access. The driver-version half is load-bearing here because the affected ranges are per-architecture (Midgard r6p0-r32p0, Bifrost r0p0-r42p0, Valhall r19p0-r42p0, Avalon r41p0-r42p0), so a fleet policy keyed on OS version alone will pass a device still running an in-range driver. The distinguishing test is to enrol a handset pinned below the bulletin level that carried the fixed driver and confirm the policy actually denies it protected-resource access — an estate that lists the stale device on a report while it keeps its mailbox has recorded the exposure rather than removed it. Precondition, stated because it is where this control is usually over-claimed: restricting installation to vetted store applications raises the bar for getting the attacker's non-privileged app onto the handset, but it does not evict an app already installed and does not cover one delivered through the normal store channel, so a device suspected of already running it belongs on the incident path, not the install-policy path. Terminal state: the entry's framework_coverage records that handsets past their Android security-update window keep this leak with no available fix, so for those models the fixed build is unreachable and the requirement ends in removing the device from the estate on a dated schedule — a rule that ends at 'every handset reports the fixed driver' marks a model whose OEM has stopped shipping bulletins as compliant while it stays exposed.",
+        "evidence": "Packet: cisa_kev true, kev_date 2023-04-07, active_exploitation 'confirmed', poc_available true; CVSS 3.3, RWEP 66; ai_discovered false. Vector: 'Memory leak vulnerability in Mali GPU Kernel Driver in Midgard GPU Kernel Driver all versions from r6p0 - r32p0, Bifrost GPU Kernel Driver all versions from r0p0 - r42p0, Valhall GPU Kernel Driver all versions from r19p0 - r42p0, and Avalon GPU Kernel Driver all versions from r41p0 - r42p0 allows a non-privileged user to make valid GPU processing operations that expose sensitive kernel metadata.' attack_vector: a non-privileged app performs valid Mali GPU operations that cause the driver to expose raw kernel pointers via a user-readable timeline-stream ring buffer, leaking kernel metadata that defeats KASLR and enables the next stage of a kernel exploit chain. patch_available true, live_patch_available false; live_patch_notes: 'No live-patch mechanism; remediation is the Arm-fixed Mali GPU driver delivered via the OEM/Android security update, applied through a device update that reboots the handset.'",
+        "gap_closes": [
+          "AU-Essential-8-Patch",
+          "ISO-27001-2022-A.8.8",
+          "NIST-800-53-SI-2",
+          "UK-CAF-B4"
+        ]
+      },
+      {
+        "id": "NEW-CTRL-001",
+        "name": "CISA-KEV-RESPONSE-SLA",
+        "description": "A CVSS 3.3 information disclosure does not survive a severity-threshold patch queue. Most estates set the intake bar at high or critical, so this entry is filtered out of the remediation backlog before a human reads it — which is exactly how a KASLR-defeat primitive stays resident on a handset fleet. Applied to this CVE, the control means the 2023-04-07 KEV listing rather than the 3.3 base score starts the clock on the Mali driver, and queue position follows the packet's RWEP of 66 with confirmed exploitation and a public PoC. The distinguishing test is to reconstruct the remediation queue as it stood on the KEV date and look for this CVE: on a CVSS-banded triage it will be absent rather than deferred, and absent is the failure this control exists to catch — a deferred item at least has an owner. Precondition, and it is the whole difficulty on a handset estate: the operator does not control when an OEM ships the Arm-fixed driver, so for a model whose bulletin has not landed the requirement cannot resolve to the patch branch at all. It resolves instead to the control's documented-compensating-control branch, and on this CVE that branch can only be an access decision, because the packet has the leak occurring during valid GPU processing operations that user space is entitled to perform — there is no on-device setting, driver option or monitoring rule that removes the read, so a compensating control recorded as 'hardened' or 'monitored' here is a compensating control that does not exist.",
+        "evidence": "Packet: cisa_kev true, kev_date 2023-04-07, active_exploitation 'confirmed', poc_available true. CVSS 3.3 against RWEP 66 — the packet's attack_vector states why the two disagree: the exposed kernel metadata is what 'defeats KASLR and enables the next stage of a kernel exploit chain'. The vector states the exposure occurs when a non-privileged user makes valid GPU processing operations. patch_available true, live_patch_available false; live_patch_notes: 'No live-patch mechanism; remediation is the Arm-fixed Mali GPU driver delivered via the OEM/Android security update, applied through a device update that reboots the handset.'",
+        "gap_closes": [
+          "ISO-27001-2022-A.8.8",
+          "NIS2-Art21-vulnerability-handling",
+          "NIST-800-53-SI-2"
+        ]
+      }
+    ]
   },
   "CVE-2022-27926": {
     "name": "Synacor Zimbra Collaboration Suite (ZCS) Cross-Site Scripting (XSS) Vulnerability (CVE-2022-27926)",