@privacyscrubber/sdk 2.4.8 → 2.4.9

package/README.md

@@ -207,7 +207,13 @@ const { scrubbedText, tokenMap } = sanitize(rawCrmPayload, {
     ],
 
     // Exclude specific domain terms or false-positive nouns (case-insensitive; also prevents multi-word NAME matches containing these terms):
-    ignoreList: ['Materialmangel', 'Störgrund', 'Prüfstand', 'Linie 3']
+    ignoreList: ['Materialmangel', 'Störgrund', 'Prüfstand', 'Linie 3'],
+
+    // Optional: Entity-type allowlist filter (e.g., ['NAME'] or ['PERSON'] to mask ONLY person names while keeping batch numbers, order IDs, and phone/date fields readable):
+    // entities: ['NAME'],
+
+    // Optional: Enable higher-recall person-name detection in free text:
+    // highRecallNames: true
 });
 ```
 

package/cli.js

@@ -32,7 +32,7 @@ for (const a of rawArgs) {
   }
 }
 
-let sdkVersion = '2.4.8';
+let sdkVersion = '2.4.9';
 try {
   const pkg = JSON.parse(fs.readFileSync(path.join(__dirname, 'package.json'), 'utf8'));
   sdkVersion = pkg.version || sdkVersion;

package/examples/python-bridge/fastapi_example.py

@@ -59,7 +59,7 @@ if FASTAPI_AVAILABLE:
     app = FastAPI(
         title="PrivacyScrubber ZTDS Sidecar Proxy",
         description="Zero-Trust Data Sanitization & In-Memory PII Shield for LLM Pipelines",
-        version="2.4.8"
+        version="2.4.9"
     )
 
     @app.on_event("startup")

package/examples/python-bridge/privacyscrubber/__init__.py

@@ -24,7 +24,7 @@ from privacyscrubber.exceptions import (
     PrivacyScrubberLicenseError,
 )
 
-__version__ = "2.4.8"
+__version__ = "2.4.9"
 __author__ = "Ilya Sibiryakov (BrandMeWeb)"
 __license__ = "Proprietary"
 

package/examples/python-bridge/privacyscrubber/bridge_daemon.cjs

@@ -115,7 +115,7 @@ rl.on('line', (line) => {
                 sendResponse(id, {
                     pong: true,
                     engine: 'PrivacyScrubber ZTDS',
-                    version: sdk.SDK_VERSION || '2.4.8',
+                    version: sdk.SDK_VERSION || '2.4.9',
                     uptime: process.uptime()
                 });
                 break;

package/examples/python-bridge/privacyscrubber/cli.py

@@ -50,7 +50,7 @@ def main() -> None:
     args = parser.parse_args()
 
     if args.version:
-        print("PrivacyScrubber ZTDS Python SDK v2.4.8")
+        print("PrivacyScrubber ZTDS Python SDK v2.4.9")
         sys.exit(0)
 
     if not args.command:

package/examples/python-bridge/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "privacyscrubber"
-version = "2.4.8"
+version = "2.4.9"
 description = "Zero-Trust Data Sanitization & In-Memory PII Shield for Python AI Pipelines (ZTDS RFC v1.0)"
 readme = "README.md"
 requires-python = ">=3.8"

package/examples/python-bridge/setup.py

@@ -13,7 +13,7 @@ with open(os.path.join(here, "README.md"), encoding="utf-8") as f:
 
 setup(
     name="privacyscrubber",
-    version="2.4.8",
+    version="2.4.9",
     description="Zero-Trust Data Sanitization & In-Memory PII Shield for Python AI Pipelines (ZTDS RFC v1.0)",
     long_description=long_description,
     long_description_content_type="text/markdown",

package/index.cjs

@@ -39,7 +39,7 @@ const SDK_VERSION = (() => {
             return pkgJson.version;
         }
     } catch (_) {}
-    return '2.4.8';
+    return '2.4.9';
 })();
 
 /**
@@ -142,8 +142,9 @@ function buildAuditTelemetry(currentTokenMap = {}, extraStats = {}) {
             if (!t) continue;
             const tokenStr = typeof t === 'string' ? t : ((t && (t.token || t.mask)) ? String(t.token || t.mask) : '');
             if (!tokenStr) continue;
-            const match = tokenStr.match(/\[([A-Z0-9_]+)_\d+\]/);
-            const baseType = match ? match[1] : (tokenStr.replace(/\[|\]/g, '').replace(/_[0-9]+$/, '') || 'CUSTOM');
+            const match = tokenStr.match(/\[([A-Za-z0-9_]+)_\d+\]/);
+            const rawBase = match ? match[1] : tokenStr.replace(/\[|\]|<|>|\{|\}|__/g, '').replace(/_[0-9]+$/, '');
+            const baseType = (rawBase.replace(/[^A-Za-z0-9_]/g, '') || 'CUSTOM').toUpperCase();
             entities[baseType] = (entities[baseType] || 0) + 1;
             totalCount++;
         }
@@ -153,7 +154,7 @@ function buildAuditTelemetry(currentTokenMap = {}, extraStats = {}) {
     let riskLevel = 'LOW EXPOSURE';
 
     if (totalCount > 0) {
-        const hasHighRisk = types.some(t => ['ID', 'SSN', 'CREDIT_CARD', 'PASSPORT', 'BANK', 'API_KEY', 'SECRET', 'PASSWORD', 'MRN', 'KEY', 'AWS_KEY', 'STRIPE_KEY', 'JWT_TOKEN', 'API_TOKEN'].includes(t));
+        const hasHighRisk = types.some(t => ['ID', 'SSN', 'CREDIT_CARD', 'PASSPORT', 'BANK', 'API_KEY', 'SECRET', 'PASSWORD', 'MRN', 'KEY', 'PHI', 'PRIVILEGE', 'AWS_KEY', 'STRIPE_KEY', 'JWT_TOKEN', 'API_TOKEN'].includes(t));
         if (hasHighRisk || types.length >= 3 || totalCount >= 10) {
             riskLevel = 'CRITICAL (HIGH EXPOSURE)';
         } else if (types.length >= 2 || totalCount >= 3) {
@@ -178,10 +179,10 @@ function buildAuditTelemetry(currentTokenMap = {}, extraStats = {}) {
         frameworksSet.add('PCI DSS v4.0');
         frameworksSet.add('IRC § 7216 & GLBA');
     }
-    if (types.some(t => ['MRN', 'HEALTH', 'MEDICAL', 'PATIENT', 'DIAGNOSIS', 'NPI', 'DEA'].includes(t))) {
+    if (types.some(t => ['PHI', 'MRN', 'HEALTH', 'MEDICAL', 'PATIENT', 'DIAGNOSIS', 'NPI', 'DEA'].includes(t))) {
         frameworksSet.add('HIPAA §164.514');
     }
-    if (types.some(t => ['CASE_NUMBER', 'DOCKET', 'PRIVILEGED', 'LEGAL', 'ATTORNEY', 'COURT', 'CONTRACT', 'NDA'].includes(t))) {
+    if (types.some(t => ['PRIVILEGE', 'CASE_NUMBER', 'DOCKET', 'PRIVILEGED', 'LEGAL', 'ATTORNEY', 'COURT', 'CONTRACT', 'NDA'].includes(t))) {
         frameworksSet.add('FRE 502 (Privilege)');
     }
     if (types.some(t => ['API_KEY', 'SECRET', 'PASSWORD', 'TOKEN', 'KEY', 'AWS_KEY', 'JWT_TOKEN', 'API_TOKEN', 'CREDENTIAL'].includes(t))) {
@@ -191,6 +192,26 @@ function buildAuditTelemetry(currentTokenMap = {}, extraStats = {}) {
         frameworksSet.add('EU AI Act (Art. 50)');
     }
 
+    const frameworksList = Array.from(frameworksSet);
+    const getFrameworkSlug = (name) => {
+        const lower = name.toLowerCase();
+        if (lower.includes('gdpr')) return 'gdpr';
+        if (lower.includes('ccpa')) return 'ccpa';
+        if (lower.includes('soc')) return 'soc2';
+        if (lower.includes('iso')) return 'iso27001';
+        if (lower.includes('pci')) return 'pci';
+        if (lower.includes('hipaa')) return 'hipaa';
+        if (lower.includes('nist')) return 'nist';
+        if (lower.includes('irc') || lower.includes('7216')) return 'irc7216';
+        if (lower.includes('fre') || lower.includes('502')) return 'fre502';
+        if (lower.includes('ai act')) return 'euaiact';
+        return lower.replace(/[^a-z0-9]/g, '');
+    };
+    const triggeredFrameworks = frameworksList.map(name => ({
+        name,
+        id: getFrameworkSlug(name)
+    }));
+
     const ipHoldings = {
         standard: 'IETF draft-sibiryakov-ztds-protocol-01',
         standardUrl: 'https://datatracker.ietf.org/doc/draft-sibiryakov-ztds-protocol/',
@@ -262,10 +283,12 @@ function buildAuditTelemetry(currentTokenMap = {}, extraStats = {}) {
 
     return {
         totalCount,
+        totalDetected: totalCount,
         entities,
         types,
         riskLevel,
-        frameworksList: Array.from(frameworksSet),
+        frameworksList,
+        triggeredFrameworks,
         auditScope,
         statutoryCitations,
         ipHoldings,
@@ -515,7 +538,7 @@ We request approval to acquire a **Developer SDK License** (\$299/month flat or
  */
 const SDK_VERTICAL_DETECTORS = [
     { profile: 'medical', label: 'Medical PHI (HIPAA)', regex: /\b(?:Patient\s*(?:Name|ID|DOB|Record)|MRN\b|Medical\s+Record|Attending\s+Physician|Chief\s+Complaint|Diagnosis\s*:|ICD-10|HIPAA|Clinical\s+Intake|Discharge\s+Summary|MedPod|NPI\s*(?:#|No|:)?\s*\d{10}|Medicare\s*(?:ID|Beneficiary|Claim)|Medicaid\s*ID|Prescription|Rx\s*#|Blood\s+Type|Vital\s+Signs|Radiology|Lab\s+Results|EHR\b|EMR\b)\b/gi },
-    { profile: 'accounting', label: 'Tax & Accounting', regex: /\b(?:Form\s*1040|Form\s*W-2|Form\s*1099|Adjusted\s+Gross\s+Income|Employer\s+Identification\s+Number|Schedule\s+C\b|Steuer-ID|USt-IdNr|Tax\s+Return|General\s+Ledger|Withholding\s+Allowance|Total\s+Tax\s+Liability)\b/gi },
+    { profile: 'accounting', label: 'Tax & Accounting', regex: /\b(?:Form\s*1040|Form\s*W-2|Form\s*1099|Adjusted\s+Gross\s+Income|Assessable\s+Income|Employer\s+Identification\s+Number|Schedule\s+C\b|Steuer-ID|USt-IdNr|Tax\s+Return|General\s+Ledger|Withholding\s+Allowance|Total\s+Tax\s+Liability|ABN\b|TFN\b|Tax\s+File\s+Number|Australian\s+Business\s+Number|Business\s+Activity\s+Statement|PAYG\s+(?:Withholding|Instalment)|Franking\s+Credits?)\b/gi },
     { profile: 'wealthmgmt', label: 'Wealth Management', regex: /\b(?:Revocable\s+Living\s+Trust|Irrevocable\s+(?:Generation-Skipping\s+)?Trust|Grantor\s*:|Settlor\s*:|Trustee\s*:|Beneficiary\s+Designation|Roth\s+IRA|Asset\s+Allocation|FBO\b|CUSIP\s*:|Grantor\s+Retained\s+Annuity\s+Trust|AUM\s*:)\b/gi },
     { profile: 'legal', label: 'Legal & Contracts', regex: /\b(?:Non-Disclosure\s+Agreement|Attorney-Client\s+Privilege|Docket\s+No|Hereinafter\s+referred|Indemnification\s+Clause|Governing\s+Law\s+and\s+Jurisdiction|Plaintiff\b|Defendant\b|Subpoena\b|Deposition\b|State\s+Bar\s*#)\b/gi },
     { profile: 'finance', label: 'Finance & Banking', regex: /\b(?:Wire\s+Transfer\s+Authorization|SWIFT\s*\/?\s*BIC\s*:|IBAN\s*:\s*[A-Z]{2}\d{2}|ABA\s+Routing|ACH\s+Transfer|PCI-DSS|Primary\s+Account\s+Number)\b/gi },
@@ -616,7 +639,9 @@ function sanitize(text, options = {}) {
         throwOnLimit = true,
         onLimit = 'throw',
         detectSecrets = false,
-        ignoreList = null
+        ignoreList = null,
+        entities = options.entityTypes || options.enabledEntities || null,
+        highRecallNames = Boolean(options.highRecallNames || options.nameRecall === 'high')
     } = options;
 
     const isAutoProfile = Boolean(
@@ -732,8 +757,23 @@ function sanitize(text, options = {}) {
         }
     }
 
+    let normalizedEntities = null;
+    if (entities) {
+        const rawArr = entities instanceof Set ? Array.from(entities) : (Array.isArray(entities) ? entities : [entities]);
+        normalizedEntities = rawArr.filter(Boolean).map(e => {
+            const u = String(e).trim().toUpperCase();
+            if (u === 'PERSON' || u === 'PERSONS' || u === 'PEOPLE' || u === 'NAMES') return 'NAME';
+            if (u === 'EMAILS') return 'EMAIL';
+            if (u === 'PHONES' || u === 'TELEPHONE') return 'PHONE';
+            if (u === 'IDS' || u === 'IDENTIFIER' || u === 'IDENTIFIERS') return 'ID';
+            if (u === 'ADDRESSES' || u === 'LOCATION' || u === 'LOCATIONS') return 'ADDRESS';
+            return u;
+        });
+        if (normalizedEntities.length === 0) normalizedEntities = null;
+    }
+
     const activeRules = [...effectiveCustomRules];
-    if (effectiveProfile.toLowerCase() === 'dev' || detectSecrets === true) {
+    if ((effectiveProfile.toLowerCase() === 'dev' || detectSecrets === true) && (!normalizedEntities || normalizedEntities.includes('SECRET') || normalizedEntities.includes('CUSTOM'))) {
         DEVOPS_SECRETS_DETECTOR.forEach(rule => {
             activeRules.push({
                 name: rule.name,
@@ -757,7 +797,9 @@ function sanitize(text, options = {}) {
         effectiveProfile,
         existingSessionMap,
         isPro,
-        effectiveIgnoreList
+        effectiveIgnoreList,
+        normalizedEntities,
+        Boolean(highRecallNames)
     );
 
     const endTime = (typeof performance !== 'undefined' && typeof performance.now === 'function') 
@@ -1085,6 +1127,8 @@ class PrivacyScrubberEngine {
         this.generateReceipt = config.generateReceipt !== false;
         this.sessionMap = { ...(config.initialSessionMap || {}) };
         this.ignoreList = new Set(config.ignoreList || []);
+        this.entities = config.entities || config.entityTypes || config.enabledEntities || null;
+        this.highRecallNames = Boolean(config.highRecallNames || config.nameRecall === 'high');
         this.maxSessionTokens = config.maxSessionTokens || 10000;
         
         this.license = validateLicense(this.licenseKey);
@@ -1126,7 +1170,9 @@ class PrivacyScrubberEngine {
             generateReceipt: options.generateReceipt !== undefined ? options.generateReceipt : this.generateReceipt,
             throwOnLimit: options.throwOnLimit !== undefined ? options.throwOnLimit : true,
             detectSecrets: options.detectSecrets !== undefined ? options.detectSecrets : false,
-            ignoreList: mergedIgnoreList.size > 0 ? mergedIgnoreList : null
+            ignoreList: mergedIgnoreList.size > 0 ? mergedIgnoreList : null,
+            entities: options.entities || options.entityTypes || options.enabledEntities || this.entities,
+            highRecallNames: options.highRecallNames !== undefined ? Boolean(options.highRecallNames) : (options.nameRecall === 'high' ? true : this.highRecallNames)
         };
 
         const result = sanitize(text, mergedOptions);
@@ -2666,13 +2712,13 @@ function buildStreamReplacer(tokenMap = {}, options = {}) {
 
     for (const k of sortedKeys) {
         const val = tokenMap[k];
-        const inner = k.replace(/^\[|<|\{\{?|__|\]|>|\}\}?|__/g, '');
+        const inner = k.replace(/^(?:\[|<|\{\{?|__|【|［|〔|「|『)|(?:\]|>|\}\}?|__|】|］|〕|」|』)$/g, '');
         const match = inner.match(/^([A-Za-z_0-9]+?)[-_ ]?(\d+)$/);
 
         if (match) {
             const label = escapeRegex(match[1]).replace(/[_-]/g, '\\\\?[_-]');
             const idx = match[2];
-            const patternStr = '(?:(?:\\[|\\\\u005[Bb]|%5[Bb]|&(?:amp;)?#0*91;|&(?:amp;)?#[xX]0*5[Bb];|&(?:amp;)?lbrack;|【|［|〔|「|『)\\s*[*`~]*' + label + '[-_\\\\\\s]*0*' + idx + '[*`~]*[.,;:!?]?\\s*(?:\\]|\\\\u005[Dd]|%5[Dd]|&(?:amp;)?#0*93;|&(?:amp;)?#[xX]0*5[Dd];|&(?:amp;)?rbrack;|】|］|〕|」|』)|<\\s*' + label + '[-_\\\\\\s]*0*' + idx + '[.,;:!?]?\\s*>|\\{\\{?\\s*' + label + '[-_\\\\\\s]*0*' + idx + '[.,;:!?]?\\s*\\}\\}?|__\\s*' + label + '[-_\\\\\\s]*0*' + idx + '\\s*__|\\b' + label + '[-_\\s]+0*' + idx + '\\b)';
+            const patternStr = '(?:(?:\\[|\\\\u005[Bb]|%5[Bb]|&(?:amp;)?#0*91;|&(?:amp;)?#[xX]0*5[Bb];|&(?:amp;)?(?:lbrack|lsqb);|【|［|〔|「|『)\\s*[*`~]*' + label + '[-_\\\\\\s]*0*' + idx + '[*`~]*[.,;:!?]?\\s*(?:\\]|\\\\u005[Dd]|%5[Dd]|&(?:amp;)?#0*93;|&(?:amp;)?#[xX]0*5[Dd];|&(?:amp;)?(?:rbrack|rsqb);|】|］|〕|」|』)|<\\s*' + label + '[-_\\\\\\s]*0*' + idx + '[.,;:!?]?\\s*>|\\{\\{?\\s*' + label + '[-_\\\\\\s]*0*' + idx + '[.,;:!?]?\\s*\\}\\}?|__\\s*' + label + '[-_\\\\\\s]*0*' + idx + '\\s*__|\\b' + label + '[-_\\s]+0*' + idx + '\\b)';
             patterns.push(patternStr);
             const normKey = (match[1] + '_' + parseInt(idx, 10)).toUpperCase();
             lookup.set(normKey, val);
@@ -2702,8 +2748,8 @@ function buildStreamReplacer(tokenMap = {}, options = {}) {
                 restored = lookup.get(coreMatch.toUpperCase());
             } else {
                 const inner = coreMatch
-                    .replace(/^(?:\[|\\u005[Bb]|%5[Bb]|&(?:amp;)?#0*91;|&(?:amp;)?#[xX]0*5[Bb];|&(?:amp;)?lbrack;|【|［|〔|「|『|<|\{\{?|__)|(?:\]|\\u005[Dd]|%5[Dd]|&(?:amp;)?#0*93;|&(?:amp;)?#[xX]0*5[Dd];|&(?:amp;)?rbrack;|】|］|〕|」|』|>|\}\}?|__)$/gi, '')
-                    .replace(/^[*`~\s]+|[.,;:!?*`~\s]+$/g, '')
+                    .replace(/^(?:\[|\\u005[Bb]|%5[Bb]|&(?:amp;)?#0*91;|&(?:amp;)?#[xX]0*5[Bb];|&(?:amp;)?(?:lbrack|lsqb);|【|［|〔|「|『|<|\{\{?|__)|(?:\]|\\u005[Dd]|%5[Dd]|&(?:amp;)?#0*93;|&(?:amp;)?#[xX]0*5[Dd];|&(?:amp;)?(?:rbrack|rsqb);|】|］|〕|」|』|>|\}\}?|__)$/gi, '')
+                    .replace(/^[*`_~\s]+|[.,;:!?*`_~\s]+$/g, '')
                     .trim();
                 const m = inner.match(/^([A-Za-z_0-9]+?)[-_\\\s]*(\d+)$/);
                 if (m) {
@@ -2816,18 +2862,37 @@ class StreamingDetokenizer {
         // Candidate unclosed token at tail of buffer (inspect tail window up to 160 chars)
         // Matches incomplete tokens starting with [, 【, ［, <, {, __, %5B, \u005b, or HTML entity &#91; / &lbrack;
         const tailLen = 160;
-        const tail = this.pendingBuffer.length > tailLen ? this.pendingBuffer.slice(-tailLen) : this.pendingBuffer;
-        const offset = this.pendingBuffer.length > tailLen ? this.pendingBuffer.length - tailLen : 0;
-        const m = tail.match(/(?:(?:\[|【|［|〔|「|『|<|\{\{?|__?|%5[Bb]|\\u005[Bb]|&(?:amp;)?(?:#0*91;|#[xX]0*5[Bb];|lbrack;))[A-Za-z0-9_\-\u0410-\u044F\u0401\u0451 *`~.\\]{0,128}(?:%5?|&(?:a(?:m(?:p(?:;)?)?)?)?(?:#(?:0*9?3?|[xX]0*5?[Dd]?)?|r(?:b(?:r(?:a(?:c(?:k)?)?)?)?)?)?)?|%[0-9A-Fa-f]?|\\(?:u(?:0(?:0(?:5)?)?)?)?|&(?:a(?:m(?:p(?:;)?)?)?)?(?:#(?:0*9?1?|[xX]0*5?)?|l(?:b(?:r(?:a(?:c(?:k)?)?)?)?)?)?)$/);
+        const tailRegex = /(?:(?:\[|【|［|〔|「|『|<|\{\{?|__|%5[Bb]|\\u005[Bb]|&(?:amp;)?(?:#0*91;|#[xX]0*5[Bb];|lbrack;))[A-Za-z0-9_\-\u0410-\u044F\u0401\u0451 *`~.\\]{0,128}(?:%5?|&(?:a(?:m(?:p(?:;)?)?)?)?(?:#(?:0*9?3?|[xX]0*5?[Dd]?)?|r(?:b(?:r(?:a(?:c(?:k)?)?)?)?)?)?)?|%[0-9A-Fa-f]?|\\(?:u(?:0(?:0(?:5)?)?)?)?|&(?:a(?:m(?:p(?:;)?)?)?)?(?:#(?:0*9?1?|[xX]0*5?)?|l(?:b(?:r(?:a(?:c(?:k)?)?)?)?)?)?|_)$/;
+        let tail = this.pendingBuffer.length > tailLen ? this.pendingBuffer.slice(-tailLen) : this.pendingBuffer;
+        let offset = this.pendingBuffer.length > tailLen ? this.pendingBuffer.length - tailLen : 0;
+        let m = tail.match(tailRegex);
 
         if (m) {
             const candidate = m[0];
             // Closed double-underscore (__NAME_1__) or JSON Unicode (\u005bNAME_1\u005d) tokens match character class and can be emitted immediately.
-            const isClosedUnderscoreOrUnicode = (candidate.length >= 4 && candidate.startsWith('__') && candidate.endsWith('__')) || /\\u005[Dd]/.test(candidate);
-            if (isClosedUnderscoreOrUnicode && this.replacer(candidate) !== candidate) {
-                const toEmit = this.pendingBuffer;
-                this.pendingBuffer = '';
-                return this.replacer(toEmit);
+            const uClose = candidate.startsWith('__') ? candidate.indexOf('__', 2) : -1;
+            const uniMatch = candidate.match(/\\u005[Dd]/);
+            const closeLen = uClose !== -1 ? uClose + 2 : (uniMatch ? uniMatch.index + 6 : -1);
+            if (closeLen !== -1) {
+                const closedToken = candidate.slice(0, closeLen);
+                if (this.replacer(closedToken) !== closedToken) {
+                    const splitAfterToken = offset + m.index + closeLen;
+                    const prefixEmitted = this.replacer(this.pendingBuffer.slice(0, splitAfterToken));
+                    this.pendingBuffer = this.pendingBuffer.slice(splitAfterToken);
+                    if (!this.pendingBuffer) return prefixEmitted;
+                    tail = this.pendingBuffer.length > tailLen ? this.pendingBuffer.slice(-tailLen) : this.pendingBuffer;
+                    offset = this.pendingBuffer.length > tailLen ? this.pendingBuffer.length - tailLen : 0;
+                    m = tail.match(tailRegex);
+                    if (!m) {
+                        const rest = this.replacer(this.pendingBuffer);
+                        this.pendingBuffer = '';
+                        return prefixEmitted + rest;
+                    }
+                    const sIdx = offset + m.index;
+                    const safe = this.pendingBuffer.slice(0, sIdx);
+                    this.pendingBuffer = this.pendingBuffer.slice(sIdx);
+                    return prefixEmitted + (safe ? this.replacer(safe) : '');
+                }
             }
             const splitIdx = offset + m.index;
             const safeToEmit = this.pendingBuffer.slice(0, splitIdx);

package/index.d.ts

@@ -74,6 +74,14 @@ export interface SanitizeOptions {
     returnFullSessionMap?: boolean;
     /** Set or array of strings or exact terms to preserve unmasked in cleartext. */
     ignoreList?: string[] | Set<string>;
+    /** Optional entity-type allowlist filter (e.g. ['NAME'] or ['PERSON'] to mask only person names while keeping batch/order numbers, dates, and IDs readable). */
+    entities?: string[];
+    /** Alias for entities. */
+    entityTypes?: string[];
+    /** Enable higher-recall person-name detection in free text (Default: false). */
+    highRecallNames?: boolean;
+    /** Alias for highRecallNames ('default' | 'high'). */
+    nameRecall?: 'default' | 'high';
     /** Whether to generate a CISO Markdown audit receipt (Default: true). */
     generateReceipt?: boolean;
     /** Throw PrivacyScrubberLicenseError when free tier limits are reached (Default: true). */

package/index.js

@@ -41,7 +41,7 @@ export const SDK_VERSION = (() => {
             return pkgJson.version;
         }
     } catch (_) {}
-    return '2.4.8';
+    return '2.4.9';
 })();
 
 /**
@@ -144,8 +144,9 @@ export function buildAuditTelemetry(currentTokenMap = {}, extraStats = {}) {
             if (!t) continue;
             const tokenStr = typeof t === 'string' ? t : ((t && (t.token || t.mask)) ? String(t.token || t.mask) : '');
             if (!tokenStr) continue;
-            const match = tokenStr.match(/\[([A-Z0-9_]+)_\d+\]/);
-            const baseType = match ? match[1] : (tokenStr.replace(/\[|\]/g, '').replace(/_[0-9]+$/, '') || 'CUSTOM');
+            const match = tokenStr.match(/\[([A-Za-z0-9_]+)_\d+\]/);
+            const rawBase = match ? match[1] : tokenStr.replace(/\[|\]|<|>|\{|\}|__/g, '').replace(/_[0-9]+$/, '');
+            const baseType = (rawBase.replace(/[^A-Za-z0-9_]/g, '') || 'CUSTOM').toUpperCase();
             entities[baseType] = (entities[baseType] || 0) + 1;
             totalCount++;
         }
@@ -155,7 +156,7 @@ export function buildAuditTelemetry(currentTokenMap = {}, extraStats = {}) {
     let riskLevel = 'LOW EXPOSURE';
 
     if (totalCount > 0) {
-        const hasHighRisk = types.some(t => ['ID', 'SSN', 'CREDIT_CARD', 'PASSPORT', 'BANK', 'API_KEY', 'SECRET', 'PASSWORD', 'MRN', 'KEY', 'AWS_KEY', 'STRIPE_KEY', 'JWT_TOKEN', 'API_TOKEN'].includes(t));
+        const hasHighRisk = types.some(t => ['ID', 'SSN', 'CREDIT_CARD', 'PASSPORT', 'BANK', 'API_KEY', 'SECRET', 'PASSWORD', 'MRN', 'KEY', 'PHI', 'PRIVILEGE', 'AWS_KEY', 'STRIPE_KEY', 'JWT_TOKEN', 'API_TOKEN'].includes(t));
         if (hasHighRisk || types.length >= 3 || totalCount >= 10) {
             riskLevel = 'CRITICAL (HIGH EXPOSURE)';
         } else if (types.length >= 2 || totalCount >= 3) {
@@ -180,10 +181,10 @@ export function buildAuditTelemetry(currentTokenMap = {}, extraStats = {}) {
         frameworksSet.add('PCI DSS v4.0');
         frameworksSet.add('IRC § 7216 & GLBA');
     }
-    if (types.some(t => ['MRN', 'HEALTH', 'MEDICAL', 'PATIENT', 'DIAGNOSIS', 'NPI', 'DEA'].includes(t))) {
+    if (types.some(t => ['PHI', 'MRN', 'HEALTH', 'MEDICAL', 'PATIENT', 'DIAGNOSIS', 'NPI', 'DEA'].includes(t))) {
         frameworksSet.add('HIPAA §164.514');
     }
-    if (types.some(t => ['CASE_NUMBER', 'DOCKET', 'PRIVILEGED', 'LEGAL', 'ATTORNEY', 'COURT', 'CONTRACT', 'NDA'].includes(t))) {
+    if (types.some(t => ['PRIVILEGE', 'CASE_NUMBER', 'DOCKET', 'PRIVILEGED', 'LEGAL', 'ATTORNEY', 'COURT', 'CONTRACT', 'NDA'].includes(t))) {
         frameworksSet.add('FRE 502 (Privilege)');
     }
     if (types.some(t => ['API_KEY', 'SECRET', 'PASSWORD', 'TOKEN', 'KEY', 'AWS_KEY', 'JWT_TOKEN', 'API_TOKEN', 'CREDENTIAL'].includes(t))) {
@@ -193,6 +194,26 @@ export function buildAuditTelemetry(currentTokenMap = {}, extraStats = {}) {
         frameworksSet.add('EU AI Act (Art. 50)');
     }
 
+    const frameworksList = Array.from(frameworksSet);
+    const getFrameworkSlug = (name) => {
+        const lower = name.toLowerCase();
+        if (lower.includes('gdpr')) return 'gdpr';
+        if (lower.includes('ccpa')) return 'ccpa';
+        if (lower.includes('soc')) return 'soc2';
+        if (lower.includes('iso')) return 'iso27001';
+        if (lower.includes('pci')) return 'pci';
+        if (lower.includes('hipaa')) return 'hipaa';
+        if (lower.includes('nist')) return 'nist';
+        if (lower.includes('irc') || lower.includes('7216')) return 'irc7216';
+        if (lower.includes('fre') || lower.includes('502')) return 'fre502';
+        if (lower.includes('ai act')) return 'euaiact';
+        return lower.replace(/[^a-z0-9]/g, '');
+    };
+    const triggeredFrameworks = frameworksList.map(name => ({
+        name,
+        id: getFrameworkSlug(name)
+    }));
+
     const ipHoldings = {
         standard: 'IETF draft-sibiryakov-ztds-protocol-01',
         standardUrl: 'https://datatracker.ietf.org/doc/draft-sibiryakov-ztds-protocol/',
@@ -264,10 +285,12 @@ export function buildAuditTelemetry(currentTokenMap = {}, extraStats = {}) {
 
     return {
         totalCount,
+        totalDetected: totalCount,
         entities,
         types,
         riskLevel,
-        frameworksList: Array.from(frameworksSet),
+        frameworksList,
+        triggeredFrameworks,
         auditScope,
         statutoryCitations,
         ipHoldings,
@@ -517,7 +540,7 @@ We request approval to acquire a **Developer SDK License** (\$299/month flat or
  */
 const SDK_VERTICAL_DETECTORS = [
     { profile: 'medical', label: 'Medical PHI (HIPAA)', regex: /\b(?:Patient\s*(?:Name|ID|DOB|Record)|MRN\b|Medical\s+Record|Attending\s+Physician|Chief\s+Complaint|Diagnosis\s*:|ICD-10|HIPAA|Clinical\s+Intake|Discharge\s+Summary|MedPod|NPI\s*(?:#|No|:)?\s*\d{10}|Medicare\s*(?:ID|Beneficiary|Claim)|Medicaid\s*ID|Prescription|Rx\s*#|Blood\s+Type|Vital\s+Signs|Radiology|Lab\s+Results|EHR\b|EMR\b)\b/gi },
-    { profile: 'accounting', label: 'Tax & Accounting', regex: /\b(?:Form\s*1040|Form\s*W-2|Form\s*1099|Adjusted\s+Gross\s+Income|Employer\s+Identification\s+Number|Schedule\s+C\b|Steuer-ID|USt-IdNr|Tax\s+Return|General\s+Ledger|Withholding\s+Allowance|Total\s+Tax\s+Liability)\b/gi },
+    { profile: 'accounting', label: 'Tax & Accounting', regex: /\b(?:Form\s*1040|Form\s*W-2|Form\s*1099|Adjusted\s+Gross\s+Income|Assessable\s+Income|Employer\s+Identification\s+Number|Schedule\s+C\b|Steuer-ID|USt-IdNr|Tax\s+Return|General\s+Ledger|Withholding\s+Allowance|Total\s+Tax\s+Liability|ABN\b|TFN\b|Tax\s+File\s+Number|Australian\s+Business\s+Number|Business\s+Activity\s+Statement|PAYG\s+(?:Withholding|Instalment)|Franking\s+Credits?)\b/gi },
     { profile: 'wealthmgmt', label: 'Wealth Management', regex: /\b(?:Revocable\s+Living\s+Trust|Irrevocable\s+(?:Generation-Skipping\s+)?Trust|Grantor\s*:|Settlor\s*:|Trustee\s*:|Beneficiary\s+Designation|Roth\s+IRA|Asset\s+Allocation|FBO\b|CUSIP\s*:|Grantor\s+Retained\s+Annuity\s+Trust|AUM\s*:)\b/gi },
     { profile: 'legal', label: 'Legal & Contracts', regex: /\b(?:Non-Disclosure\s+Agreement|Attorney-Client\s+Privilege|Docket\s+No|Hereinafter\s+referred|Indemnification\s+Clause|Governing\s+Law\s+and\s+Jurisdiction|Plaintiff\b|Defendant\b|Subpoena\b|Deposition\b|State\s+Bar\s*#)\b/gi },
     { profile: 'finance', label: 'Finance & Banking', regex: /\b(?:Wire\s+Transfer\s+Authorization|SWIFT\s*\/?\s*BIC\s*:|IBAN\s*:\s*[A-Z]{2}\d{2}|ABA\s+Routing|ACH\s+Transfer|PCI-DSS|Primary\s+Account\s+Number)\b/gi },
@@ -618,7 +641,9 @@ export function sanitize(text, options = {}) {
         throwOnLimit = true,
         onLimit = 'throw',
         detectSecrets = false,
-        ignoreList = null
+        ignoreList = null,
+        entities = options.entityTypes || options.enabledEntities || null,
+        highRecallNames = Boolean(options.highRecallNames || options.nameRecall === 'high')
     } = options;
 
     const isAutoProfile = Boolean(
@@ -734,8 +759,23 @@ export function sanitize(text, options = {}) {
         }
     }
 
+    let normalizedEntities = null;
+    if (entities) {
+        const rawArr = entities instanceof Set ? Array.from(entities) : (Array.isArray(entities) ? entities : [entities]);
+        normalizedEntities = rawArr.filter(Boolean).map(e => {
+            const u = String(e).trim().toUpperCase();
+            if (u === 'PERSON' || u === 'PERSONS' || u === 'PEOPLE' || u === 'NAMES') return 'NAME';
+            if (u === 'EMAILS') return 'EMAIL';
+            if (u === 'PHONES' || u === 'TELEPHONE') return 'PHONE';
+            if (u === 'IDS' || u === 'IDENTIFIER' || u === 'IDENTIFIERS') return 'ID';
+            if (u === 'ADDRESSES' || u === 'LOCATION' || u === 'LOCATIONS') return 'ADDRESS';
+            return u;
+        });
+        if (normalizedEntities.length === 0) normalizedEntities = null;
+    }
+
     const activeRules = [...effectiveCustomRules];
-    if (effectiveProfile.toLowerCase() === 'dev' || detectSecrets === true) {
+    if ((effectiveProfile.toLowerCase() === 'dev' || detectSecrets === true) && (!normalizedEntities || normalizedEntities.includes('SECRET') || normalizedEntities.includes('CUSTOM'))) {
         DEVOPS_SECRETS_DETECTOR.forEach(rule => {
             activeRules.push({
                 name: rule.name,
@@ -759,7 +799,9 @@ export function sanitize(text, options = {}) {
         effectiveProfile,
         existingSessionMap,
         isPro,
-        effectiveIgnoreList
+        effectiveIgnoreList,
+        normalizedEntities,
+        Boolean(highRecallNames)
     );
 
     const endTime = (typeof performance !== 'undefined' && typeof performance.now === 'function') 
@@ -1087,6 +1129,8 @@ export class PrivacyScrubberEngine {
         this.generateReceipt = config.generateReceipt !== false;
         this.sessionMap = { ...(config.initialSessionMap || {}) };
         this.ignoreList = new Set(config.ignoreList || []);
+        this.entities = config.entities || config.entityTypes || config.enabledEntities || null;
+        this.highRecallNames = Boolean(config.highRecallNames || config.nameRecall === 'high');
         this.maxSessionTokens = config.maxSessionTokens || 10000;
         
         this.license = validateLicense(this.licenseKey);
@@ -1128,7 +1172,9 @@ export class PrivacyScrubberEngine {
             generateReceipt: options.generateReceipt !== undefined ? options.generateReceipt : this.generateReceipt,
             throwOnLimit: options.throwOnLimit !== undefined ? options.throwOnLimit : true,
             detectSecrets: options.detectSecrets !== undefined ? options.detectSecrets : false,
-            ignoreList: mergedIgnoreList.size > 0 ? mergedIgnoreList : null
+            ignoreList: mergedIgnoreList.size > 0 ? mergedIgnoreList : null,
+            entities: options.entities || options.entityTypes || options.enabledEntities || this.entities,
+            highRecallNames: options.highRecallNames !== undefined ? Boolean(options.highRecallNames) : (options.nameRecall === 'high' ? true : this.highRecallNames)
         };
 
         const result = sanitize(text, mergedOptions);
@@ -2668,13 +2714,13 @@ export function buildStreamReplacer(tokenMap = {}, options = {}) {
 
     for (const k of sortedKeys) {
         const val = tokenMap[k];
-        const inner = k.replace(/^\[|<|\{\{?|__|\]|>|\}\}?|__/g, '');
+        const inner = k.replace(/^(?:\[|<|\{\{?|__|【|［|〔|「|『)|(?:\]|>|\}\}?|__|】|］|〕|」|』)$/g, '');
         const match = inner.match(/^([A-Za-z_0-9]+?)[-_ ]?(\d+)$/);
 
         if (match) {
             const label = escapeRegex(match[1]).replace(/[_-]/g, '\\\\?[_-]');
             const idx = match[2];
-            const patternStr = '(?:(?:\\[|\\\\u005[Bb]|%5[Bb]|&(?:amp;)?#0*91;|&(?:amp;)?#[xX]0*5[Bb];|&(?:amp;)?lbrack;|【|［|〔|「|『)\\s*[*`~]*' + label + '[-_\\\\\\s]*0*' + idx + '[*`~]*[.,;:!?]?\\s*(?:\\]|\\\\u005[Dd]|%5[Dd]|&(?:amp;)?#0*93;|&(?:amp;)?#[xX]0*5[Dd];|&(?:amp;)?rbrack;|】|］|〕|」|』)|<\\s*' + label + '[-_\\\\\\s]*0*' + idx + '[.,;:!?]?\\s*>|\\{\\{?\\s*' + label + '[-_\\\\\\s]*0*' + idx + '[.,;:!?]?\\s*\\}\\}?|__\\s*' + label + '[-_\\\\\\s]*0*' + idx + '\\s*__|\\b' + label + '[-_\\s]+0*' + idx + '\\b)';
+            const patternStr = '(?:(?:\\[|\\\\u005[Bb]|%5[Bb]|&(?:amp;)?#0*91;|&(?:amp;)?#[xX]0*5[Bb];|&(?:amp;)?(?:lbrack|lsqb);|【|［|〔|「|『)\\s*[*`~]*' + label + '[-_\\\\\\s]*0*' + idx + '[*`~]*[.,;:!?]?\\s*(?:\\]|\\\\u005[Dd]|%5[Dd]|&(?:amp;)?#0*93;|&(?:amp;)?#[xX]0*5[Dd];|&(?:amp;)?(?:rbrack|rsqb);|】|］|〕|」|』)|<\\s*' + label + '[-_\\\\\\s]*0*' + idx + '[.,;:!?]?\\s*>|\\{\\{?\\s*' + label + '[-_\\\\\\s]*0*' + idx + '[.,;:!?]?\\s*\\}\\}?|__\\s*' + label + '[-_\\\\\\s]*0*' + idx + '\\s*__|\\b' + label + '[-_\\s]+0*' + idx + '\\b)';
             patterns.push(patternStr);
             const normKey = (match[1] + '_' + parseInt(idx, 10)).toUpperCase();
             lookup.set(normKey, val);
@@ -2704,8 +2750,8 @@ export function buildStreamReplacer(tokenMap = {}, options = {}) {
                 restored = lookup.get(coreMatch.toUpperCase());
             } else {
                 const inner = coreMatch
-                    .replace(/^(?:\[|\\u005[Bb]|%5[Bb]|&(?:amp;)?#0*91;|&(?:amp;)?#[xX]0*5[Bb];|&(?:amp;)?lbrack;|【|［|〔|「|『|<|\{\{?|__)|(?:\]|\\u005[Dd]|%5[Dd]|&(?:amp;)?#0*93;|&(?:amp;)?#[xX]0*5[Dd];|&(?:amp;)?rbrack;|】|］|〕|」|』|>|\}\}?|__)$/gi, '')
-                    .replace(/^[*`~\s]+|[.,;:!?*`~\s]+$/g, '')
+                    .replace(/^(?:\[|\\u005[Bb]|%5[Bb]|&(?:amp;)?#0*91;|&(?:amp;)?#[xX]0*5[Bb];|&(?:amp;)?(?:lbrack|lsqb);|【|［|〔|「|『|<|\{\{?|__)|(?:\]|\\u005[Dd]|%5[Dd]|&(?:amp;)?#0*93;|&(?:amp;)?#[xX]0*5[Dd];|&(?:amp;)?(?:rbrack|rsqb);|】|］|〕|」|』|>|\}\}?|__)$/gi, '')
+                    .replace(/^[*`_~\s]+|[.,;:!?*`_~\s]+$/g, '')
                     .trim();
                 const m = inner.match(/^([A-Za-z_0-9]+?)[-_\\\s]*(\d+)$/);
                 if (m) {
@@ -2818,18 +2864,37 @@ export class StreamingDetokenizer {
         // Candidate unclosed token at tail of buffer (inspect tail window up to 160 chars)
         // Matches incomplete tokens starting with [, 【, ［, <, {, __, %5B, \u005b, or HTML entity &#91; / &lbrack;
         const tailLen = 160;
-        const tail = this.pendingBuffer.length > tailLen ? this.pendingBuffer.slice(-tailLen) : this.pendingBuffer;
-        const offset = this.pendingBuffer.length > tailLen ? this.pendingBuffer.length - tailLen : 0;
-        const m = tail.match(/(?:(?:\[|【|［|〔|「|『|<|\{\{?|__?|%5[Bb]|\\u005[Bb]|&(?:amp;)?(?:#0*91;|#[xX]0*5[Bb];|lbrack;))[A-Za-z0-9_\-\u0410-\u044F\u0401\u0451 *`~.\\]{0,128}(?:%5?|&(?:a(?:m(?:p(?:;)?)?)?)?(?:#(?:0*9?3?|[xX]0*5?[Dd]?)?|r(?:b(?:r(?:a(?:c(?:k)?)?)?)?)?)?)?|%[0-9A-Fa-f]?|\\(?:u(?:0(?:0(?:5)?)?)?)?|&(?:a(?:m(?:p(?:;)?)?)?)?(?:#(?:0*9?1?|[xX]0*5?)?|l(?:b(?:r(?:a(?:c(?:k)?)?)?)?)?)?)$/);
+        const tailRegex = /(?:(?:\[|【|［|〔|「|『|<|\{\{?|__|%5[Bb]|\\u005[Bb]|&(?:amp;)?(?:#0*91;|#[xX]0*5[Bb];|lbrack;))[A-Za-z0-9_\-\u0410-\u044F\u0401\u0451 *`~.\\]{0,128}(?:%5?|&(?:a(?:m(?:p(?:;)?)?)?)?(?:#(?:0*9?3?|[xX]0*5?[Dd]?)?|r(?:b(?:r(?:a(?:c(?:k)?)?)?)?)?)?)?|%[0-9A-Fa-f]?|\\(?:u(?:0(?:0(?:5)?)?)?)?|&(?:a(?:m(?:p(?:;)?)?)?)?(?:#(?:0*9?1?|[xX]0*5?)?|l(?:b(?:r(?:a(?:c(?:k)?)?)?)?)?)?|_)$/;
+        let tail = this.pendingBuffer.length > tailLen ? this.pendingBuffer.slice(-tailLen) : this.pendingBuffer;
+        let offset = this.pendingBuffer.length > tailLen ? this.pendingBuffer.length - tailLen : 0;
+        let m = tail.match(tailRegex);
 
         if (m) {
             const candidate = m[0];
             // Closed double-underscore (__NAME_1__) or JSON Unicode (\u005bNAME_1\u005d) tokens match character class and can be emitted immediately.
-            const isClosedUnderscoreOrUnicode = (candidate.length >= 4 && candidate.startsWith('__') && candidate.endsWith('__')) || /\\u005[Dd]/.test(candidate);
-            if (isClosedUnderscoreOrUnicode && this.replacer(candidate) !== candidate) {
-                const toEmit = this.pendingBuffer;
-                this.pendingBuffer = '';
-                return this.replacer(toEmit);
+            const uClose = candidate.startsWith('__') ? candidate.indexOf('__', 2) : -1;
+            const uniMatch = candidate.match(/\\u005[Dd]/);
+            const closeLen = uClose !== -1 ? uClose + 2 : (uniMatch ? uniMatch.index + 6 : -1);
+            if (closeLen !== -1) {
+                const closedToken = candidate.slice(0, closeLen);
+                if (this.replacer(closedToken) !== closedToken) {
+                    const splitAfterToken = offset + m.index + closeLen;
+                    const prefixEmitted = this.replacer(this.pendingBuffer.slice(0, splitAfterToken));
+                    this.pendingBuffer = this.pendingBuffer.slice(splitAfterToken);
+                    if (!this.pendingBuffer) return prefixEmitted;
+                    tail = this.pendingBuffer.length > tailLen ? this.pendingBuffer.slice(-tailLen) : this.pendingBuffer;
+                    offset = this.pendingBuffer.length > tailLen ? this.pendingBuffer.length - tailLen : 0;
+                    m = tail.match(tailRegex);
+                    if (!m) {
+                        const rest = this.replacer(this.pendingBuffer);
+                        this.pendingBuffer = '';
+                        return prefixEmitted + rest;
+                    }
+                    const sIdx = offset + m.index;
+                    const safe = this.pendingBuffer.slice(0, sIdx);
+                    this.pendingBuffer = this.pendingBuffer.slice(sIdx);
+                    return prefixEmitted + (safe ? this.replacer(safe) : '');
+                }
             }
             const splitIdx = offset + m.index;
             const safeToEmit = this.pendingBuffer.slice(0, splitIdx);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@privacyscrubber/sdk",
-  "version": "2.4.8",
+  "version": "2.4.9",
   "description": "In-process consumer privacy defense and pre-emptive PII interception SDK for Node.js, Next.js, and AI pipelines. Zero data loss, zero third parties, sub-millisecond in-memory execution.",
   "main": "index.cjs",
   "module": "index.js",