modern-pdf-lib 0.34.0 → 0.35.0

package/dist/{src-PxZwqIKQ.cjs → src-C-mBizNU.cjs}

@@ -1,4 +1,4 @@
-const require_pdfDocument = require("./pdfDocument-Dyj4AmT1.cjs");
+const require_pdfDocument = require("./pdfDocument-D5qz7duP.cjs");
 const require_pdfObjects = require("./pdfObjects-BcPlSI0a.cjs");
 const require_streamDecode = require("./streamDecode-kQ-REV7v.cjs");
 const require_pdfForm = require("./pdfForm-BVS_do95.cjs");
@@ -277,7 +277,7 @@ function compressStream(stream, level) {
 * @returns              The incremental save result.
 */
 async function saveDocumentIncremental(originalBytes, doc, options) {
-	const { buildDocumentStructure } = await Promise.resolve().then(() => require("./pdfDocument-Dyj4AmT1.cjs")).then((n) => n.pdfCatalog_exports);
+	const { buildDocumentStructure } = await Promise.resolve().then(() => require("./pdfDocument-D5qz7duP.cjs")).then((n) => n.pdfCatalog_exports);
 	const registry = doc.getRegistry();
 	const structure = buildDocumentStructure(doc.getInternalPages().map((p) => p.finalize()), {
 		producer: doc.getProducer(),
@@ -28081,6 +28081,1041 @@ function buildCertPath(leafCertDer, intermediates, anchors) {
 	};
 }
 //#endregion
+//#region src/security/threatScanner.ts
+/** Numeric rank for severities so we can take a maximum. */
+const SEVERITY_RANK = {
+	low: 1,
+	medium: 2,
+	high: 3
+};
+/** Reverse lookup from rank back to severity. */
+const RANK_SEVERITY = [
+	"low",
+	"medium",
+	"high"
+];
+/**
+* File-name extensions that denote directly executable / script payloads.
+* An attachment ending in one of these is treated as high severity because a
+* viewer's "open attachment" (or a `/Launch` action targeting it) can run it.
+*/
+const EXECUTABLE_EXTENSIONS = /* @__PURE__ */ new Set([
+	"exe",
+	"com",
+	"scr",
+	"pif",
+	"cpl",
+	"msi",
+	"msp",
+	"mst",
+	"dll",
+	"sys",
+	"drv",
+	"ocx",
+	"bat",
+	"cmd",
+	"ps1",
+	"psm1",
+	"vbs",
+	"vbe",
+	"vb",
+	"wsf",
+	"wsh",
+	"js",
+	"jse",
+	"jar",
+	"class",
+	"sh",
+	"bash",
+	"zsh",
+	"ksh",
+	"csh",
+	"command",
+	"app",
+	"dmg",
+	"pkg",
+	"lnk",
+	"url",
+	"reg",
+	"hta",
+	"gadget",
+	"inf",
+	"scf",
+	"apk",
+	"deb",
+	"rpm",
+	"run",
+	"bin",
+	"elf",
+	"so",
+	"dylib",
+	"py",
+	"pyc",
+	"pyw",
+	"rb",
+	"pl",
+	"php"
+]);
+/** Read a `PdfName` value (without the leading `/`) from a dict entry. */
+function nameValue$1(obj) {
+	if (obj !== void 0 && obj.kind === "name") {
+		const v = obj.value;
+		return v.startsWith("/") ? v.slice(1) : v;
+	}
+}
+/** Read a `PdfString` value from a dict entry. */
+function stringValue(obj) {
+	if (obj !== void 0 && obj.kind === "string") return obj.value;
+}
+/** Format a `PdfRef` as the canonical `"N G R"` string, if available. */
+function refString(ref) {
+	return ref ? `${ref.objectNumber} ${ref.generationNumber} R` : void 0;
+}
+/** Lower-case file extension of a name, or `''` when there is none. */
+function extensionOf(name) {
+	const base = name.replace(/[\\/]+/g, "/").split("/").pop() ?? name;
+	const dot = base.lastIndexOf(".");
+	if (dot < 0 || dot === base.length - 1) return "";
+	return base.slice(dot + 1).toLowerCase();
+}
+/**
+* Scan a PDF for hostile constructs and return a {@link ThreatReport}.
+*
+* The scan never executes any document content. It parses the PDF and walks
+* the parsed object graph (see the module header for the precise method and
+* ISO 32000 references). When parsing fails, a guarded raw-byte fallback runs.
+*
+* @param pdf  Raw PDF bytes.
+* @returns    A report of findings plus the aggregate `riskLevel`.
+*/
+async function scanPdfThreats(pdf) {
+	const findings = [];
+	let scannedGraph = false;
+	try {
+		scanObjectGraph((await require_pdfDocument.loadPdf(pdf, { ignoreEncryption: true })).getRegistry(), findings);
+		scannedGraph = true;
+	} catch {
+		scannedGraph = false;
+	}
+	if (!scannedGraph) scanRawBytesFallback(pdf, findings);
+	return {
+		findings,
+		riskLevel: aggregateRisk(findings)
+	};
+}
+/**
+* Walk every registered object and emit findings.
+*
+* We iterate the registry (every resolved indirect object). For each
+* dictionary (including stream dictionaries) we inspect that dict's *own
+* keys* and, where relevant, the resolved action-subtype `/S` *name*. Content
+* streams' raw bytes are never substring-scanned here, which is what makes the
+* scan precise (drawn text that merely says "JavaScript" is not a finding).
+*/
+function scanObjectGraph(registry, findings) {
+	for (const entry of registry) {
+		const obj = entry.object;
+		const refStr = refString(entry.ref);
+		let dict;
+		if (obj.kind === "dict") dict = obj;
+		else if (obj.kind === "stream") dict = obj.dict;
+		if (dict === void 0) continue;
+		inspectDict(dict, refStr, findings);
+	}
+}
+/**
+* Inspect a single dictionary's keys/values for dangerous constructs.
+*
+* @param dict     The dictionary (or stream dictionary).
+* @param refStr   The owning object's `"N G R"` string, if known.
+* @param findings Accumulator.
+*/
+function inspectDict(dict, refStr, findings) {
+	if (dict.has("/OpenAction")) findings.push({
+		category: "OpenAction",
+		severity: "medium",
+		detail: "Document defines an /OpenAction that runs automatically when the PDF is opened (ISO 32000 §12.6.3).",
+		objectRef: refStr
+	});
+	if (dict.has("/AA")) findings.push({
+		category: "AA",
+		severity: "medium",
+		detail: "Document or object defines /AA additional actions that fire on viewer events such as open/close/print/save (ISO 32000 §12.6.2).",
+		objectRef: refStr
+	});
+	if (dict.has("/JavaScript")) findings.push({
+		category: "JavaScript",
+		severity: "high",
+		detail: "A /JavaScript name tree (catalog /Names → /JavaScript) is present; these scripts execute automatically on open (ISO 32000 §12.6.4.16, §7.7.4).",
+		objectRef: refStr
+	});
+	const subtype = nameValue$1(dict.get("/S"));
+	if (subtype !== void 0) inspectActionSubtype(subtype, dict, refStr, findings);
+	if (dict.has("/XFA")) findings.push({
+		category: "XFA",
+		severity: "low",
+		detail: "Document contains an XFA form (/AcroForm → /XFA, ISO 32000 §12.7.8); XFA carries its own scripting and data-binding model.",
+		objectRef: refStr
+	});
+	const annotSubtype = nameValue$1(dict.get("/Subtype"));
+	if (annotSubtype === "RichMedia") findings.push({
+		category: "RichMedia",
+		severity: "low",
+		detail: "Rich-media annotation (/Subtype /RichMedia, ISO 32000-2 §13.6) embeds interactive multimedia (e.g. Flash/video).",
+		objectRef: refStr
+	});
+	else if (annotSubtype === "Movie") findings.push({
+		category: "Movie",
+		severity: "low",
+		detail: "Movie annotation (/Subtype /Movie, ISO 32000 §13.4) embeds video/animation content.",
+		objectRef: refStr
+	});
+	else if (annotSubtype === "Sound") findings.push({
+		category: "Sound",
+		severity: "low",
+		detail: "Sound annotation (/Subtype /Sound, ISO 32000 §13.3) embeds audio content.",
+		objectRef: refStr
+	});
+	if (dict.has("/RichMedia") && annotSubtype !== "RichMedia") findings.push({
+		category: "RichMedia",
+		severity: "low",
+		detail: "Object references rich-media content via a /RichMedia entry (ISO 32000-2 §13.6).",
+		objectRef: refStr
+	});
+	if (dict.has("/Movie") && annotSubtype !== "Movie") findings.push({
+		category: "Movie",
+		severity: "low",
+		detail: "Object references movie content via a /Movie entry (ISO 32000 §13.4).",
+		objectRef: refStr
+	});
+	if (dict.has("/Sound") && annotSubtype !== "Sound" && subtype !== "Sound") findings.push({
+		category: "Sound",
+		severity: "low",
+		detail: "Object references sound content via a /Sound entry (ISO 32000 §13.3).",
+		objectRef: refStr
+	});
+	if (nameValue$1(dict.get("/Type")) === "Filespec" || dict.has("/EF")) {
+		const fileName = stringValue(dict.get("/UF")) ?? stringValue(dict.get("/F"));
+		if (fileName !== void 0) {
+			const ext = extensionOf(fileName);
+			if (ext !== "" && EXECUTABLE_EXTENSIONS.has(ext)) findings.push({
+				category: "EmbeddedFile",
+				severity: "high",
+				detail: `Embedded file "${fileName}" has an executable/script extension ".${ext}" (ISO 32000 §7.11.3/§7.11.4); opening it can run code.`,
+				objectRef: refStr
+			});
+		}
+	}
+}
+/**
+* Flag a dangerous action `/S` subtype. Subtypes are ISO 32000 §12.6.4 /
+* §12.7.5 action types.
+*/
+function inspectActionSubtype(subtype, dict, refStr, findings) {
+	switch (subtype) {
+		case "JavaScript":
+			findings.push({
+				category: "JavaScript",
+				severity: "high",
+				detail: "JavaScript action (/S /JavaScript, ISO 32000 §12.6.4.16) executes script in the viewer.",
+				objectRef: refStr
+			});
+			break;
+		case "Launch":
+			findings.push({
+				category: "Launch",
+				severity: "high",
+				detail: "Launch action (/S /Launch, ISO 32000 §12.6.4.5) can run an external application or open a file.",
+				objectRef: refStr
+			});
+			break;
+		case "URI":
+			findings.push({
+				category: "URI",
+				severity: "medium",
+				detail: `URI action (/S /URI, ISO 32000 §12.6.4.7) opens a remote URL ${describeUri(dict)}.`,
+				objectRef: refStr
+			});
+			break;
+		case "SubmitForm":
+			findings.push({
+				category: "SubmitForm",
+				severity: "medium",
+				detail: "SubmitForm action (/S /SubmitForm, ISO 32000 §12.7.5.2) sends form data to a URL (potential data exfiltration).",
+				objectRef: refStr
+			});
+			break;
+		case "ImportData":
+			findings.push({
+				category: "ImportData",
+				severity: "low",
+				detail: "ImportData action (/S /ImportData, ISO 32000 §12.7.5.4) imports external form data.",
+				objectRef: refStr
+			});
+			break;
+		case "GoToR":
+			findings.push({
+				category: "GoToR",
+				severity: "low",
+				detail: "Remote Go-To action (/S /GoToR, ISO 32000 §12.6.4.3) targets an external document.",
+				objectRef: refStr
+			});
+			break;
+		case "Movie":
+			findings.push({
+				category: "Movie",
+				severity: "low",
+				detail: "Movie action (/S /Movie, ISO 32000 §12.6.4.10) plays embedded video.",
+				objectRef: refStr
+			});
+			break;
+		case "Sound":
+			findings.push({
+				category: "Sound",
+				severity: "low",
+				detail: "Sound action (/S /Sound, ISO 32000 §12.6.4.8) plays embedded audio.",
+				objectRef: refStr
+			});
+			break;
+		default: break;
+	}
+}
+/** Append the target URL of a /URI action to the detail string, if present. */
+function describeUri(dict) {
+	const uri = stringValue(dict.get("/URI"));
+	return uri !== void 0 ? `→ "${uri}"` : "";
+}
+/**
+* Best-effort raw-byte fallback used only when the PDF cannot be parsed.
+*
+* To stay precise we match constructs only in their *structural* dictionary
+* form — action subtypes as `/S /Type` (optionally without the inner space)
+* and `/OpenAction`/`/AA`/`/JavaScript` as dictionary keys (a `/` immediately
+* followed by the token and then a PDF delimiter). This avoids flagging the
+* same words when they appear as content-stream text or inside a literal
+* string. It is heuristic and clearly less reliable than the graph walk.
+*/
+function scanRawBytesFallback(pdf, findings) {
+	const text = new TextDecoder("latin1").decode(pdf);
+	const keyForm = (key) => new RegExp(`/${key}(?=[\\s/(<\\[\\]>]|$)`);
+	if (keyForm("OpenAction").test(text)) findings.push({
+		category: "OpenAction",
+		severity: "medium",
+		detail: "Raw scan: /OpenAction key present (ISO 32000 §12.6.3)."
+	});
+	if (keyForm("AA").test(text)) findings.push({
+		category: "AA",
+		severity: "medium",
+		detail: "Raw scan: /AA additional-actions key present (ISO 32000 §12.6.2)."
+	});
+	if (keyForm("JavaScript").test(text) || /\/S\s*\/JavaScript\b/.test(text)) findings.push({
+		category: "JavaScript",
+		severity: "high",
+		detail: "Raw scan: JavaScript action/name tree present (ISO 32000 §12.6.4.16)."
+	});
+	if (/\/S\s*\/Launch\b/.test(text)) findings.push({
+		category: "Launch",
+		severity: "high",
+		detail: "Raw scan: /S /Launch action present (ISO 32000 §12.6.4.5)."
+	});
+	if (/\/S\s*\/URI\b/.test(text)) findings.push({
+		category: "URI",
+		severity: "medium",
+		detail: "Raw scan: /S /URI action present (ISO 32000 §12.6.4.7)."
+	});
+	if (/\/S\s*\/SubmitForm\b/.test(text)) findings.push({
+		category: "SubmitForm",
+		severity: "medium",
+		detail: "Raw scan: /S /SubmitForm action present (ISO 32000 §12.7.5.2)."
+	});
+	if (/\/S\s*\/ImportData\b/.test(text)) findings.push({
+		category: "ImportData",
+		severity: "low",
+		detail: "Raw scan: /S /ImportData action present (ISO 32000 §12.7.5.4)."
+	});
+	if (/\/S\s*\/GoToR\b/.test(text)) findings.push({
+		category: "GoToR",
+		severity: "low",
+		detail: "Raw scan: /S /GoToR remote go-to present (ISO 32000 §12.6.4.3)."
+	});
+	if (/\/Subtype\s*\/RichMedia\b/.test(text) || keyForm("RichMedia").test(text)) findings.push({
+		category: "RichMedia",
+		severity: "low",
+		detail: "Raw scan: RichMedia annotation present (ISO 32000-2 §13.6)."
+	});
+	if (/\/Subtype\s*\/Movie\b/.test(text) || /\/S\s*\/Movie\b/.test(text)) findings.push({
+		category: "Movie",
+		severity: "low",
+		detail: "Raw scan: Movie content/action present (ISO 32000 §13.4)."
+	});
+	if (/\/Subtype\s*\/Sound\b/.test(text) || /\/S\s*\/Sound\b/.test(text)) findings.push({
+		category: "Sound",
+		severity: "low",
+		detail: "Raw scan: Sound content/action present (ISO 32000 §13.3)."
+	});
+	if (keyForm("XFA").test(text)) findings.push({
+		category: "XFA",
+		severity: "low",
+		detail: "Raw scan: /XFA form present (ISO 32000 §12.7.8)."
+	});
+}
+/** Maximum severity across findings, or `'none'` when there are none. */
+function aggregateRisk(findings) {
+	let maxRank = 0;
+	for (const f of findings) {
+		const rank = SEVERITY_RANK[f.severity];
+		if (rank > maxRank) maxRank = rank;
+	}
+	if (maxRank === 0) return "none";
+	return RANK_SEVERITY[maxRank - 1];
+}
+//#endregion
+//#region src/security/sanitize.ts
+/**
+* @module security/sanitize
+*
+* PDF sanitizer — produce a cleaned copy of a PDF with active / hidden
+* content neutralised.
+*
+* The sanitizer removes four classes of potentially dangerous or
+* privacy-leaking content from a document, each independently toggleable:
+*
+* - **javascript** — document-level JavaScript: the catalog's
+*   `/Names → /JavaScript` name tree (ISO 32000-1 §7.7.4 / §12.6.4.16)
+*   and the catalog's `/AA` (document additional-actions) dictionary
+*   (§12.6.3) plus any page-level `/AA`.
+* - **openActions** — the catalog's auto-run `/OpenAction` (§12.3.4 / §12.6.4).
+* - **embeddedFiles** — embedded file attachments: the catalog's
+*   `/Names → /EmbeddedFiles` name tree (§7.7.4 / §7.11.4) and the
+*   PDF 2.0 catalog `/AF` (associated files) array (§7.11.4).
+* - **metadata** — the XMP `/Metadata` stream (§14.3.2) and the document
+*   information `/Info` dictionary (§14.3.3).
+*
+* Implementation strategy (verified against this library's parser/writer):
+* the cleaned bytes are produced by mutating the *parsed object graph* in
+* place and re-serializing the registry directly through
+* {@link serializePdf}, **without** rebuilding the catalog. This is what
+* lets a disabled option genuinely preserve its content — the original
+* catalog (minus the removed references) is written verbatim.
+*
+* Removing a reference from the catalog is not sufficient on its own: the
+* underlying object (a JavaScript action dictionary, an `/EmbeddedFile`
+* stream, an XMP metadata stream, …) would otherwise remain physically
+* present in the output as an orphan. To guarantee the payload bytes are
+* actually gone, after deleting references this module runs the registry's
+* reachability filter from the catalog + info roots, dropping every object
+* no longer referenced.
+*
+* @packageDocumentation
+*/
+/**
+* Resolve a {@link PdfObject} that may be an indirect reference to its
+* underlying value using the registry. Non-refs are returned as-is.
+* Returns `undefined` if a ref cannot be resolved.
+*/
+function resolve$2(obj, registry) {
+	if (obj === void 0) return void 0;
+	if (obj instanceof require_pdfObjects.PdfRef) return registry.resolve(obj);
+	return obj;
+}
+/** Resolve `obj` and return it only if it is a {@link PdfDict}. */
+function asDict(obj, registry) {
+	const resolved = resolve$2(obj, registry);
+	return resolved instanceof require_pdfObjects.PdfDict ? resolved : void 0;
+}
+/** Read a {@link PdfString} value, resolving an indirect reference if needed. */
+function asStringValue(obj, registry) {
+	const resolved = resolve$2(obj, registry);
+	return resolved instanceof require_pdfObjects.PdfString ? resolved.value : void 0;
+}
+/**
+* Collect every distinct `/Type /Catalog` dictionary registered in the
+* document, paired with its reference. A well-formed PDF has exactly one,
+* but malformed or incrementally-updated files can register more than one;
+* sanitizing all of them is the conservative, correct choice. The first
+* entry is treated as the document root.
+*/
+function findCatalogEntries(registry) {
+	const byObjNum = /* @__PURE__ */ new Map();
+	for (const { ref, object } of registry) if (object instanceof require_pdfObjects.PdfDict) {
+		const type = object.get("/Type");
+		if (type instanceof require_pdfObjects.PdfName && type.value === "/Catalog") {
+			if (!byObjNum.has(ref.objectNumber)) byObjNum.set(ref.objectNumber, {
+				ref,
+				dict: object
+			});
+		}
+	}
+	return [...byObjNum.values()];
+}
+/** Collect every distinct `/Type /Page` dictionary in the registry. */
+function findPages(registry) {
+	const pages = /* @__PURE__ */ new Set();
+	for (const { object } of registry) if (object instanceof require_pdfObjects.PdfDict) {
+		const type = object.get("/Type");
+		if (type instanceof require_pdfObjects.PdfName && type.value === "/Page") pages.add(object);
+	}
+	return [...pages];
+}
+/**
+* Determine whether a PDF name tree (ISO 32000-1 §7.7.4) contains at least
+* one entry. A name-tree node holds either a `/Names` array of
+* `[key value key value …]` pairs (leaf) or a `/Kids` array of child node
+* references (intermediate). The tree is walked recursively with a visited
+* guard against cyclic references in malformed files.
+*/
+function nameTreeHasEntries(node, registry, visited = /* @__PURE__ */ new Set()) {
+	const dict = asDict(node, registry);
+	if (dict === void 0 || visited.has(dict)) return false;
+	visited.add(dict);
+	const names = resolve$2(dict.get("/Names"), registry);
+	if (names instanceof require_pdfObjects.PdfArray && names.length >= 2) return true;
+	const kids = resolve$2(dict.get("/Kids"), registry);
+	if (kids instanceof require_pdfObjects.PdfArray) {
+		for (const kid of kids.items) if (nameTreeHasEntries(kid, registry, visited)) return true;
+	}
+	return false;
+}
+/** Return the catalog's resolved `/Names` sub-dictionary, if present. */
+function getNamesDict(catalog, registry) {
+	return asDict(catalog.get("/Names"), registry);
+}
+/**
+* Determine whether an additional-actions (`/AA`) dictionary contains any
+* action whose `/S` (action type) is `/JavaScript`. Each value in an `/AA`
+* dictionary is itself an action dictionary (§12.6.3).
+*/
+function aaHasJavaScript(aa, registry) {
+	if (aa === void 0) return false;
+	for (const [, value] of aa) {
+		const action = asDict(value, registry);
+		if (action === void 0) continue;
+		const s = action.get("/S");
+		if (s instanceof require_pdfObjects.PdfName && s.value === "/JavaScript") return true;
+	}
+	return false;
+}
+/**
+* Descriptive `/Info` keys that constitute identifying metadata. The
+* auto-generated `/Producer`, `/CreationDate`, and `/ModDate` are NOT
+* treated as "present" for reporting purposes: a freshly created document
+* carries those by default and must still be considered clean. (When
+* metadata removal runs, the entire `/Info` dict is dropped regardless.)
+*/
+const DESCRIPTIVE_INFO_KEYS = [
+	"/Title",
+	"/Author",
+	"/Subject",
+	"/Keywords",
+	"/Creator"
+];
+/**
+* Locate the trailer's `/Info` dictionary in the registry. The parser does
+* not expose the trailer publicly, so the Info dict is identified
+* structurally: it is a dictionary that is NOT a catalog/page/pages node
+* and carries at least one recognised document-information key.
+*/
+function findInfoDict(registry, catalogs) {
+	const catalogDicts = new Set(catalogs.map((c) => c.dict));
+	const INFO_KEYS = [
+		"/Title",
+		"/Author",
+		"/Subject",
+		"/Keywords",
+		"/Creator",
+		"/Producer",
+		"/CreationDate",
+		"/ModDate",
+		"/Trapped"
+	];
+	for (const { object } of registry) {
+		if (!(object instanceof require_pdfObjects.PdfDict) || catalogDicts.has(object)) continue;
+		const type = object.get("/Type");
+		if (type instanceof require_pdfObjects.PdfName) {
+			const v = type.value;
+			if (v === "/Catalog" || v === "/Pages" || v === "/Page" || v === "/Font") continue;
+		}
+		if (INFO_KEYS.some((k) => object.has(k))) return object;
+	}
+}
+/** Read descriptive metadata from an `/Info` dict into a typed bag. */
+function readInfoMetadata(info, registry) {
+	const meta = {};
+	if (info === void 0) return meta;
+	const title = asStringValue(info.get("/Title"), registry);
+	if (title !== void 0) meta.title = title;
+	const author = asStringValue(info.get("/Author"), registry);
+	if (author !== void 0) meta.author = author;
+	const subject = asStringValue(info.get("/Subject"), registry);
+	if (subject !== void 0) meta.subject = subject;
+	const keywords = asStringValue(info.get("/Keywords"), registry);
+	if (keywords !== void 0) meta.keywords = keywords;
+	const creator = asStringValue(info.get("/Creator"), registry);
+	if (creator !== void 0) meta.creator = creator;
+	const producer = asStringValue(info.get("/Producer"), registry);
+	if (producer !== void 0) meta.producer = producer;
+	return meta;
+}
+/**
+* Produce a cleaned copy of a PDF with active / hidden content neutralised.
+*
+* Loads the PDF, removes each enabled class of content from the parsed
+* object graph, prunes the orphaned objects, re-serializes the cleaned
+* document, and returns the new bytes alongside a report listing the
+* classes that were actually present and removed.
+*
+* @param pdf      The source PDF bytes.
+* @param options  Which classes to remove (each defaults to `true`).
+* @returns        The cleaned PDF bytes and a {@link SanitizeReport}.
+*
+* @example
+* ```ts
+* const { pdf, report } = await sanitizePdf(bytes);
+* // report.removed === ['javascript', 'embeddedFiles', 'metadata']
+* ```
+*/
+async function sanitizePdf(pdf, options) {
+	const removeJavaScript = options?.javascript ?? true;
+	const removeOpenActions = options?.openActions ?? true;
+	const removeEmbeddedFiles = options?.embeddedFiles ?? true;
+	const removeMetadata = options?.metadata ?? true;
+	const registry = (await require_pdfDocument.loadPdf(pdf)).getRegistry();
+	const catalogs = findCatalogEntries(registry);
+	if (catalogs.length === 0) return {
+		pdf,
+		report: { removed: [] }
+	};
+	const pages = findPages(registry);
+	const infoDict = findInfoDict(registry, catalogs);
+	const removed = /* @__PURE__ */ new Set();
+	if (removeJavaScript) {
+		for (const { dict: catalog } of catalogs) {
+			const names = getNamesDict(catalog, registry);
+			if (names !== void 0 && nameTreeHasEntries(names.get("/JavaScript"), registry)) removed.add("javascript");
+			if (aaHasJavaScript(asDict(catalog.get("/AA"), registry), registry)) removed.add("javascript");
+		}
+		for (const page of pages) if (aaHasJavaScript(asDict(page.get("/AA"), registry), registry)) removed.add("javascript");
+		for (const { dict: catalog } of catalogs) {
+			const names = getNamesDict(catalog, registry);
+			if (names !== void 0) {
+				names.delete("/JavaScript");
+				if (names.size === 0) catalog.delete("/Names");
+			}
+			catalog.delete("/AA");
+		}
+		for (const page of pages) page.delete("/AA");
+	}
+	if (removeOpenActions) {
+		for (const { dict: catalog } of catalogs) if (catalog.has("/OpenAction")) {
+			removed.add("openActions");
+			catalog.delete("/OpenAction");
+		}
+	}
+	if (removeEmbeddedFiles) {
+		for (const { dict: catalog } of catalogs) {
+			const names = getNamesDict(catalog, registry);
+			if (names !== void 0 && nameTreeHasEntries(names.get("/EmbeddedFiles"), registry)) removed.add("embeddedFiles");
+			const af = resolve$2(catalog.get("/AF"), registry);
+			if (af instanceof require_pdfObjects.PdfArray && af.length > 0) removed.add("embeddedFiles");
+		}
+		for (const { dict: catalog } of catalogs) {
+			const names = getNamesDict(catalog, registry);
+			if (names !== void 0) {
+				names.delete("/EmbeddedFiles");
+				if (names.size === 0) catalog.delete("/Names");
+			}
+			catalog.delete("/AF");
+		}
+	}
+	const preservedMeta = readInfoMetadata(infoDict, registry);
+	if (removeMetadata) {
+		let metadataPresent = false;
+		for (const { dict: catalog } of catalogs) if (resolve$2(catalog.get("/Metadata"), registry) instanceof require_pdfObjects.PdfStream) metadataPresent = true;
+		if (infoDict !== void 0 && DESCRIPTIVE_INFO_KEYS.some((k) => infoDict.has(k))) metadataPresent = true;
+		if (metadataPresent) removed.add("metadata");
+		for (const { dict: catalog } of catalogs) catalog.delete("/Metadata");
+		if (infoDict !== void 0) for (const [key] of [...infoDict]) infoDict.delete(key);
+	}
+	const root = catalogs[0];
+	const pagesRef = root.dict.get("/Pages");
+	let infoRef;
+	if (removeMetadata) infoRef = require_pdfDocument.buildInfoDict({}, registry);
+	else infoRef = require_pdfDocument.buildInfoDict(preservedMeta, registry);
+	const structure = {
+		catalogRef: root.ref,
+		infoRef,
+		pagesRef: pagesRef instanceof require_pdfObjects.PdfRef ? pagesRef : root.ref
+	};
+	registry.filterReachable([structure.catalogRef, structure.infoRef]);
+	return {
+		pdf: await require_pdfDocument.serializePdf(registry, structure, { compress: true }),
+		report: { removed: [...removed] }
+	};
+}
+//#endregion
+//#region src/security/redactionVerifier.ts
+/**
+* @module security/redactionVerifier
+*
+* Detect FAILED / fake redactions — text that is still extractable underneath
+* a redaction region.
+*
+* A *secure* redaction physically removes the underlying text-showing
+* operators from the page content stream (see `redactRegions` in
+* `../render/redactContent.js`), so the redacted text cannot be recovered by
+* copy/paste, text extraction, or raw stream inspection.  A *fake* redaction
+* merely paints an opaque black box on top of the page while the original text
+* remains in the byte stream — a genuine, well-documented security failure
+* (the "select-all-and-copy" leak).  This verifier finds those leaks.
+*
+* ## How it works
+*
+* For each page named by a region:
+* 1. The page's content stream is decoded
+*    (`PdfPage.getContentStream()`) and parsed (`parseContentStream`).
+* 2. `extractTextWithPositions` replays the text/graphics-state machine and
+*    yields one positioned {@link import('../parser/textExtractor.js').TextItem}
+*    per shown text run, with its page-space origin and size.
+* 3. Every text item whose bounding box intersects a region rectangle is
+*    reported as a leak.  `clean` is true iff there are no leaks.
+*
+* ## Coordinate / encoding conventions (confirmed, not assumed)
+*
+* - **Space:** PDF user space — the coordinate space produced by
+*   `extractTextWithPositions`.  Origin is **bottom-left**, the y-axis points
+*   **up**, and units are **PDF points** (1/72 inch).  This matches
+*   `RedactRect` in `../render/redactContent.js` ("page space, PDF points,
+*   y-up, lower-left origin"), so a {@link RedactionRegion} uses exactly the
+*   same convention: `(x, y)` is the lower-left corner and `width` / `height`
+*   extend in the +x / +y directions.
+* - **Glyph box:** a `TextItem` reports `(x, y)` = the text origin
+*   (baseline-left), `width` extending +x, and `height` (equal to the font
+*   size) extending +y.  The run's axis-aligned box is therefore
+*   `[x, y, x + width, y + height]`.  Note `width`/`height` are heuristic
+*   estimates from `extractTextWithPositions` (0.5·fontSize per character;
+*   height = fontSize) because per-glyph font metrics are not available there;
+*   intersection testing tolerates this by treating any overlap as a leak.
+* - **Text decoding:** strings are decoded by `extractTextWithPositions` using
+*   the font's `/ToUnicode` CMap when present, else WinAnsiEncoding (1-byte) or
+*   2-byte Identity-H for CID fonts — see `../parser/textExtractor.js`.
+*
+* ## Region auto-detection is intentionally NOT performed
+*
+* The `regions` argument is **required**.  Deriving candidate regions from the
+* PDF itself is not reliable through the public API and is not fabricated here:
+*   - Filled black rectangles in the content stream are indistinguishable from
+*     legitimate black-filled graphics (chart bars, rules, design elements), so
+*     a heuristic would produce false positives/negatives.
+*   - `/Redact` annotations (ISO 32000-2 §12.5.6.23) *do* carry explicit
+*     `/Rect` / `/QuadPoints` regions, but a loaded page's original annotation
+*     dictionaries are not exposed through a stable public `PdfPage` accessor,
+*     so reading them here is not possible without reaching into private state.
+* Callers must therefore pass the regions they redacted (or believe were
+* redacted).  When `regions` is omitted or empty, {@link verifyRedactions}
+* throws.
+*
+* @packageDocumentation
+*/
+/**
+* Axis-aligned bounding box of a text run in page space (bottom-left origin).
+* `width`/`height` may be zero (empty text); such runs never intersect.
+*/
+function itemBox(item) {
+	const x0 = item.x;
+	const y0 = item.y;
+	const x1 = item.x + item.width;
+	const y1 = item.y + item.height;
+	return {
+		minX: Math.min(x0, x1),
+		minY: Math.min(y0, y1),
+		maxX: Math.max(x0, x1),
+		maxY: Math.max(y0, y1)
+	};
+}
+/**
+* Whether a text run's box overlaps a region rectangle.
+*
+* Uses the standard separating-axis test for axis-aligned rectangles: they
+* overlap unless one is entirely to the left/right/above/below the other.
+* Touching edges count as overlap (inclusive bounds), matching the intersection
+* test in `../render/redactContent.js`.
+*/
+function intersects$1(box, region) {
+	const rMinX = region.x;
+	const rMinY = region.y;
+	const rMaxX = region.x + region.width;
+	const rMaxY = region.y + region.height;
+	return box.minX <= rMaxX && box.maxX >= rMinX && box.minY <= rMaxY && box.maxY >= rMinY;
+}
+/**
+* Verify that the given regions of a PDF contain no still-extractable text,
+* detecting fake / failed redactions.
+*
+* For each region, the corresponding page's text is extracted with positions
+* and any text run whose bounding box intersects the region is reported as a
+* {@link RedactionLeak}.  A clean result means no text was found under any
+* region (the redactions truly removed the content, or there was none to begin
+* with).
+*
+* @param pdf      The PDF file bytes to inspect.
+* @param regions  The regions to check (REQUIRED — see the module docs for why
+*                 automatic region detection is not performed).  Coordinates are
+*                 in PDF user space: origin bottom-left, y-up, units in points;
+*                 `(x, y)` is the lower-left corner.
+* @returns        A report listing every leak; `clean` is true iff there are
+*                 none.
+* @throws         `TypeError` if `regions` is omitted or empty.
+*
+* @example
+* ```ts
+* const report = await verifyRedactions(pdfBytes, [
+*   { page: 0, x: 45, y: 545, width: 120, height: 25 },
+* ]);
+* if (!report.clean) {
+*   for (const leak of report.leaks) {
+*     console.warn(`Leak on page ${leak.page}: "${leak.text}"`);
+*   }
+* }
+* ```
+*/
+async function verifyRedactions(pdf, regions) {
+	if (regions === void 0 || regions.length === 0) throw new TypeError("verifyRedactions requires an explicit, non-empty `regions` array. Automatic redaction-region detection is not performed: black-filled rectangles are indistinguishable from legitimate graphics, and a loaded page's original /Redact annotations are not exposed through a stable public API. Pass the regions you redacted (PDF points, bottom-left origin, y-up).");
+	const doc = await require_pdfDocument.loadPdf(pdf);
+	const pageCount = doc.getPageCount();
+	const itemsByPage = /* @__PURE__ */ new Map();
+	const getItems = (pageIndex) => {
+		const cached = itemsByPage.get(pageIndex);
+		if (cached !== void 0) return cached;
+		let items = [];
+		if (pageIndex >= 0 && pageIndex < pageCount) {
+			const page = doc.getPage(pageIndex);
+			items = require_compressionAnalysis.extractTextWithPositions(require_compressionAnalysis.parseContentStream(page.getContentStream()), page.getOriginalResources());
+		}
+		itemsByPage.set(pageIndex, items);
+		return items;
+	};
+	const leaks = [];
+	for (const region of regions) {
+		if (region.width <= 0 || region.height <= 0) continue;
+		const items = getItems(region.page);
+		for (const item of items) {
+			if (item.text.length === 0) continue;
+			const box = itemBox(item);
+			if (box.maxX === box.minX && box.maxY === box.minY) continue;
+			if (intersects$1(box, region)) leaks.push({
+				page: region.page,
+				text: item.text,
+				x: item.x,
+				y: item.y
+			});
+		}
+	}
+	return {
+		leaks,
+		clean: leaks.length === 0,
+		regionsChecked: regions.length
+	};
+}
+//#endregion
+//#region src/security/encryptionInspector.ts
+const BIT_PRINT = 4;
+const BIT_MODIFY = 8;
+const BIT_COPY = 16;
+const BIT_ANNOTATE = 32;
+const BIT_FILL_FORMS = 256;
+const BIT_ASSEMBLE = 1024;
+const BIT_PRINT_HIGH_RES = 2048;
+/**
+* Inspect a PDF's encryption + permission posture.
+*
+* @param pdf  The raw PDF file bytes.
+* @returns    A {@link EncryptionReport}.  If the trailer has no
+*             `/Encrypt` entry (or the file is unparseable) the report is
+*             `{ encrypted: false }`.
+*/
+async function inspectEncryption(pdf) {
+	const located = await locateEncryptDict(pdf);
+	if (located === void 0) return { encrypted: false };
+	const { encryptDict, fileId, encryptRef } = located;
+	const filterName = getName(encryptDict, "/Filter");
+	let handler;
+	if (filterName === "/Standard") handler = "password";
+	else if (filterName === "/Adobe.PubSec") handler = "publicKey";
+	const version = getNumber(encryptDict, "/V");
+	const revision = getNumber(encryptDict, "/R");
+	const { method, keyBits } = decodeAlgorithm(encryptDict, version);
+	let permissions;
+	const pValue = getNumber(encryptDict, "/P");
+	if (handler === "password" && pValue !== void 0) permissions = decodePermissionBits(pValue);
+	let emptyUserPassword;
+	if (handler === "password") emptyUserPassword = await testEmptyUserPassword(encryptDict, fileId);
+	const report = { encrypted: true };
+	if (method !== void 0) report.method = method;
+	if (keyBits !== void 0) report.keyBits = keyBits;
+	if (version !== void 0) report.version = version;
+	if (revision !== void 0) report.revision = revision;
+	if (handler !== void 0) report.handler = handler;
+	if (emptyUserPassword !== void 0) report.emptyUserPassword = emptyUserPassword;
+	if (permissions !== void 0) report.permissions = permissions;
+	return report;
+}
+/**
+* Parse just enough of the PDF (header + xref + trailer) to resolve the
+* `/Encrypt` dictionary and the first element of the `/ID` array.
+*
+* Returns `undefined` when the file has no `/Encrypt` entry or when it
+* cannot be parsed far enough to find one.
+*/
+async function locateEncryptDict(pdf) {
+	if (pdf.length < 8) return void 0;
+	let encryptRef;
+	let fileIdFirst;
+	let entry;
+	let objectParser;
+	try {
+		objectParser = new require_pdfDocument.PdfObjectParser(new require_pdfDocument.PdfLexer(pdf), new require_pdfObjects.PdfObjectRegistry());
+		const { entries, trailer } = await new require_pdfDocument.XrefParser(pdf, objectParser).parseXref();
+		encryptRef = trailer.encryptRef;
+		if (encryptRef === void 0) return void 0;
+		fileIdFirst = trailer.id?.[0] ?? /* @__PURE__ */ new Uint8Array(0);
+		entry = entries.get(encryptRef.objectNumber);
+	} catch {
+		return;
+	}
+	if (entry === void 0 || entry.type !== "in-use") return void 0;
+	let encryptObj;
+	try {
+		const { object } = objectParser.parseIndirectObjectAt(entry.offset);
+		encryptObj = object;
+	} catch {
+		return;
+	}
+	if (encryptObj.kind !== "dict") return void 0;
+	return {
+		encryptDict: encryptObj,
+		encryptRef,
+		fileId: fileIdFirst
+	};
+}
+/**
+* Decode the bulk cipher and key length from an encryption dictionary.
+*
+* - V=1: 40-bit RC4 (Table 20/21).
+* - V=2: RC4 with `/Length` bits.
+* - V=4: inspect the crypt filter `/CFM` of the filter named by `/StmF`
+*   (falling back to `/StdCF`): `/V2` → RC4, `/AESV2` → 128-bit AES,
+*   `/AESV3` → 256-bit AES, `/Identity`/`/None` → no encryption.
+* - V=5: 256-bit AES (`/AESV3`).
+*/
+function decodeAlgorithm(dict, version) {
+	const lengthBits = getNumber(dict, "/Length");
+	if (version === 1) return {
+		method: "rc4",
+		keyBits: 40
+	};
+	if (version === 2) return {
+		method: "rc4",
+		keyBits: lengthBits ?? 40
+	};
+	if (version === 5) return {
+		method: "aes",
+		keyBits: 256
+	};
+	if (version === 4) {
+		const cfm = resolveCryptFilterMethod(dict);
+		if (cfm === "/AESV3") return {
+			method: "aes",
+			keyBits: 256
+		};
+		if (cfm === "/AESV2") return {
+			method: "aes",
+			keyBits: lengthBits ?? 128
+		};
+		if (cfm === "/V2") return {
+			method: "rc4",
+			keyBits: lengthBits ?? 128
+		};
+		return {
+			method: void 0,
+			keyBits: lengthBits ?? void 0
+		};
+	}
+	return {
+		method: void 0,
+		keyBits: lengthBits ?? void 0
+	};
+}
+/**
+* For V=4 documents, resolve the `/CFM` (crypt filter method) name of the
+* stream crypt filter.  Looks up `/StmF` in `/CF`, then falls back to the
+* conventional `/StdCF` entry.
+*/
+function resolveCryptFilterMethod(dict) {
+	const cf = dict.get("/CF");
+	if (cf === void 0 || cf.kind !== "dict") return void 0;
+	const cfDict = cf;
+	let filterName = getName(dict, "/StmF");
+	if (filterName === "/Identity") return "/Identity";
+	let stdCf = filterName !== void 0 ? cfDict.get(filterName) : void 0;
+	if (stdCf === void 0 || stdCf.kind !== "dict") {
+		stdCf = cfDict.get("/StdCF");
+		filterName = "/StdCF";
+	}
+	if (stdCf === void 0 || stdCf.kind !== "dict") return void 0;
+	return getName(stdCf, "/CFM");
+}
+/**
+* Decode the 32-bit `/P` integer into individual permission flags.
+*
+* The `/P` value is a *signed* 32-bit integer; bitwise AND in JavaScript
+* operates on the 32-bit two's-complement representation, so masks apply
+* correctly even when `/P` is negative (the usual case, since reserved
+* high bits are set to 1).
+*/
+function decodePermissionBits(p) {
+	const has = (mask) => (p & mask) !== 0;
+	const copy = has(BIT_COPY);
+	return {
+		print: has(BIT_PRINT),
+		modify: has(BIT_MODIFY),
+		copy,
+		annotate: has(BIT_ANNOTATE),
+		fillForms: has(BIT_FILL_FORMS),
+		extract: copy,
+		assemble: has(BIT_ASSEMBLE),
+		printHighRes: has(BIT_PRINT_HIGH_RES)
+	};
+}
+/**
+* Test whether the empty user password (`""`) authenticates against the
+* standard security handler.  Reuses {@link PdfEncryptionHandler.fromEncryptDict},
+* which throws when the password is incorrect; a successful return means
+* the empty password derived a valid file key.
+*
+* Returns `undefined` when the test cannot be performed (e.g. the handler
+* implementation does not support the document's /V/R).
+*/
+async function testEmptyUserPassword(dict, fileId) {
+	try {
+		await require_pdfDocument.PdfEncryptionHandler.fromEncryptDict(dict, fileId, "");
+		return true;
+	} catch (err) {
+		const message = err instanceof Error ? err.message : String(err);
+		if (/password/i.test(message)) return false;
+		return;
+	}
+}
+/** Read a `/Name` value (including the leading `/`) from a dict. */
+function getName(dict, key) {
+	const obj = dict.get(key);
+	if (obj !== void 0 && obj.kind === "name") return obj.value;
+}
+/** Read a numeric value from a dict. */
+function getNumber(dict, key) {
+	const obj = dict.get(key);
+	if (obj !== void 0 && obj.kind === "number") return obj.value;
+}
+//#endregion
 //#region src/render/matrix.ts
 /** The identity transform. */
 function identity() {
@@ -31889,15 +32924,15 @@ async function initWasm(options) {
 	if (wasmInitialized) return;
 	const inits = [];
 	if (options.deflate || options.deflateWasm) inits.push((async () => {
-		const { initDeflateWasm } = await Promise.resolve().then(() => require("./pdfDocument-Dyj4AmT1.cjs")).then((n) => n.libdeflateWasm_exports);
+		const { initDeflateWasm } = await Promise.resolve().then(() => require("./pdfDocument-D5qz7duP.cjs")).then((n) => n.libdeflateWasm_exports);
 		await initDeflateWasm(options.deflateWasm);
 	})());
 	if (options.png || options.pngWasm) inits.push((async () => {
-		const { initPngWasm } = await Promise.resolve().then(() => require("./pdfDocument-Dyj4AmT1.cjs")).then((n) => n.pngEmbed_exports);
+		const { initPngWasm } = await Promise.resolve().then(() => require("./pdfDocument-D5qz7duP.cjs")).then((n) => n.pngEmbed_exports);
 		await initPngWasm(options.pngWasm);
 	})());
 	if (options.fonts || options.fontWasm) inits.push((async () => {
-		const { initSubsetWasm } = await Promise.resolve().then(() => require("./pdfDocument-Dyj4AmT1.cjs")).then((n) => n.fontSubset_exports);
+		const { initSubsetWasm } = await Promise.resolve().then(() => require("./pdfDocument-D5qz7duP.cjs")).then((n) => n.fontSubset_exports);
 		await initSubsetWasm(options.fontWasm);
 	})());
 	if (options.jpeg || options.jpegWasm) inits.push((async () => {
@@ -33312,6 +34347,12 @@ Object.defineProperty(exports, "injectJpegMetadata", {
 		return injectJpegMetadata;
 	}
 });
+Object.defineProperty(exports, "inspectEncryption", {
+	enumerable: true,
+	get: function() {
+		return inspectEncryption;
+	}
+});
 Object.defineProperty(exports, "interpretContentStream", {
 	enumerable: true,
 	get: function() {
@@ -33684,6 +34725,12 @@ Object.defineProperty(exports, "sampleShadingColor", {
 		return sampleShadingColor;
 	}
 });
+Object.defineProperty(exports, "sanitizePdf", {
+	enumerable: true,
+	get: function() {
+		return sanitizePdf;
+	}
+});
 Object.defineProperty(exports, "saveDocumentIncremental", {
 	enumerable: true,
 	get: function() {
@@ -33702,6 +34749,12 @@ Object.defineProperty(exports, "saveIncrementalWithSignaturePreservation", {
 		return saveIncrementalWithSignaturePreservation;
 	}
 });
+Object.defineProperty(exports, "scanPdfThreats", {
+	enumerable: true,
+	get: function() {
+		return scanPdfThreats;
+	}
+});
 Object.defineProperty(exports, "searchTextItems", {
 	enumerable: true,
 	get: function() {
@@ -33978,6 +35031,12 @@ Object.defineProperty(exports, "verifyOfflineRevocation", {
 		return verifyOfflineRevocation;
 	}
 });
+Object.defineProperty(exports, "verifyRedactions", {
+	enumerable: true,
+	get: function() {
+		return verifyRedactions;
+	}
+});
 Object.defineProperty(exports, "verifySignatureDetailed", {
 	enumerable: true,
 	get: function() {
@@ -34003,4 +35062,4 @@ Object.defineProperty(exports, "wrapText", {
 	}
 });
 
-//# sourceMappingURL=src-PxZwqIKQ.cjs.map
+//# sourceMappingURL=src-C-mBizNU.cjs.map