spamscanner 7.0.7 → 7.0.9

package/dist/esm/index.js

@@ -2102,11 +2102,12 @@ function gatewayTarget(url) {
   }
   if (host === "urldefense.com" && url.pathname.startsWith("/v3/__")) {
     const rest = url.href.slice(url.href.indexOf("/v3/__") + 6);
-    const end = rest.indexOf("__;");
+    const end = rest.lastIndexOf("__;");
     return end === -1 ? null : rest.slice(0, end);
   }
   return null;
 }
+var MAX_GATEWAYS = 16;
 function unwrapLink(address2) {
   let url;
   try {
@@ -2114,8 +2115,11 @@ function unwrapLink(address2) {
   } catch {
     return null;
   }
-  for (let depth = 0; depth < 3; depth++) {
+  for (let depth = 0; depth < MAX_GATEWAYS; depth++) {
     const target = gatewayTarget(url);
+    if (!target) {
+      break;
+    }
     let next;
     try {
       next = new URL(target);
@@ -3429,8 +3433,13 @@ function oleHasMacros(buffer) {
   return false;
 }
 var PDF_ACTIVE = linear(String.raw`\/(?:JavaScript|JS|Launch)\b`);
-var PDF_ESCAPED_NAME = /\/[^\s/<>[\]()%{}]*#[\dA-Fa-f]{2}[^\s/<>[\]()%{}]*/g;
-var decodePdfNames = (text) => text.replaceAll(PDF_ESCAPED_NAME, (name) => name.replaceAll(/#([\dA-Fa-f]{2})/g, (_, hex) => String.fromCodePoint(Number.parseInt(hex, 16))));
+var escaped = (character) => {
+  const hex = character.codePointAt(0).toString(16);
+  return `(?:${character}|#${hex[0]}[${hex[1].toLowerCase()}${hex[1].toUpperCase()}])`;
+};
+var WORD_ESCAPE = String.raw`#(?:3\d|4[1-9a-fA-F]|5[\daA]|5[fF]|6[1-9a-fA-F]|7[\daA])`;
+var PDF_ACTIVE_ESCAPED = new RegExp(`/(?:${["JavaScript", "JS", "Launch"].map((name) => [...name].map((character) => escaped(character)).join("")).join("|")})(?![\\w]|${WORD_ESCAPE})`);
+var pdfActive = (text) => PDF_ACTIVE.test(text) || PDF_ACTIVE_ESCAPED.test(text);
 function inspectAttachments(attachments = []) {
   const findings = [];
   for (const attachment of attachments) {
@@ -3488,7 +3497,7 @@ function inspectAttachments(attachments = []) {
         add("macro", `Attachment "${name}" is an Office document with a VBA macro project`, { extension });
       }
     } else if (detected?.type === "pdf" || extension === "pdf") {
-      if (PDF_ACTIVE.test(decodePdfNames(content.toString("latin1")))) {
+      if (pdfActive(content.toString("latin1"))) {
         add("pdf_active", `PDF "${name}" contains JavaScript or a launch action`, { extension: "pdf" });
       }
     } else if (detected?.type === "rtf" && /\\object\b|\\objdata\b|\\objupdate\b/.test(content.toString("latin1"))) {
@@ -3536,7 +3545,7 @@ function boundAddressField(line, room) {
   const limit = Math.min(line.length, MAX_ADDRESS_FIELD, room);
   let end = start;
   let expecting = "";
-  let escaped = false;
+  let escaped2 = false;
   let separators = 0;
   let groups = 0;
   let index = start;
@@ -3545,8 +3554,8 @@ function boundAddressField(line, room) {
     if ((chr === "," || chr === ";") && ++separators > MAX_ADDRESS_SEPARATORS) {
       break;
     }
-    if (escaped) {
-      escaped = false;
+    if (escaped2) {
+      escaped2 = false;
     } else if (expecting !== "" && chr === expecting) {
       expecting = "";
       end = chr === ";" ? index + 1 : end;
@@ -3554,7 +3563,7 @@ function boundAddressField(line, room) {
       expecting = ADDRESS_OPERATORS.get(chr);
       end = chr === "," || chr === ";" ? index : end;
     } else if (expecting === '"' && chr === "\\") {
-      escaped = true;
+      escaped2 = true;
     }
     if (chr === ":" && expecting === ";" && ++groups > MAX_ADDRESS_GROUPS) {
       break;
@@ -3697,9 +3706,14 @@ var SIGNATURE_FIELDS = /* @__PURE__ */ new Set(["dkim-signature", "arc-message-s
 var AUTH_FIELDS = /* @__PURE__ */ new Set([...KEPT_FIELDS, "arc-seal", "arc-message-signature"]);
 var MAX_SIGNATURES = 10;
 var MAX_DKIM_SIGNATURES = 8;
+function isAttachedMessage(attachment) {
+  const type = String(attachment.contentType).toLowerCase();
+  return type === "message/rfc822" || type === "message/global" || /\.(?:eml|u8msg)$/i.test(String(attachment.filename));
+}
 var MAX_NESTED_DEPTH = 3;
 var MAX_ATTACHMENTS = 100;
 var MAX_NESTED_BYTES = 25 * 1024 * 1024;
+var MAX_ATTACHED_MESSAGES = 100;
 var PARSE_TIMEOUT = 6e4;
 async function simpleParse(raw, { timeout = PARSE_TIMEOUT, parser = simpleParser } = {}) {
   const input = Readable.from([finishable(unfoldHeaders(raw))], { objectMode: false, autoDestroy: false });
@@ -3757,18 +3771,18 @@ function headerValue2(line) {
   let key = null;
   let inKey = false;
   let quoted = false;
-  let escaped = false;
+  let escaped2 = false;
   let value = "";
   for (const chr of line.slice(line.indexOf(":") + 1).replaceAll(/\r?\n[ \t]*/g, " ")) {
     if (inKey && chr === "=") {
       key = value.trim().toLowerCase();
       inKey = false;
       value = "";
-    } else if (inKey || escaped) {
+    } else if (inKey || escaped2) {
       value += chr;
-      escaped = false;
+      escaped2 = false;
     } else if (chr === "\\") {
-      escaped = true;
+      escaped2 = true;
     } else if (chr === '"') {
       quoted = !quoted;
     } else if (chr === ";" && !quoted) {
@@ -3816,7 +3830,7 @@ function unfoldHeaders(raw) {
     if (blank && type.boundary) {
       boundaries.push(Buffer2.from(type.boundary));
     }
-    head = blank && type.value === "message/rfc822" && ["", "7bit", "8bit", "binary"].includes(encoding) && first.get("content-disposition")?.value === "inline";
+    head = blank && ["message/rfc822", "message/global"].includes(type.value) && ["", "7bit", "8bit", "binary"].includes(encoding) && first.get("content-disposition")?.value === "inline";
   };
   for (let start = 0; start < raw.length; ) {
     const lf = raw.indexOf(10, start);
@@ -3947,25 +3961,45 @@ async function parseWithin(raw, parse = parseMessage, deadline = Date.now() + PA
     }
   }
 }
-async function expandAttachments(attachments = [], errors = [], parse = parseMessage) {
+async function expandAttachments(attachments = [], errors = [], parse = parseMessage, unopened = []) {
   const all = [];
   const deadline = Date.now() + PARSE_TIMEOUT;
   let budget = MAX_NESTED_BYTES;
   let nested = 0;
+  let tooMany = false;
+  let opened = 0;
   const visit = async (list, prefix, depth) => {
     for (const attachment of list) {
       if (prefix) {
         if (nested >= MAX_ATTACHMENTS) {
+          if (!tooMany) {
+            tooMany = true;
+            unopened.push({ filename: prefix, reason: `more than ${MAX_ATTACHMENTS} attachments in attached messages` });
+          }
           return;
         }
         nested++;
       }
       const name = attachment.filename || (prefix ? "unnamed attachment" : "");
+      const filename = prefix ? `${prefix}/${name}` : name || "attached message";
       all.push(prefix ? { ...attachment, filename: `${prefix}/${name}` } : attachment);
       const { content } = attachment;
-      const attached = attachment.contentType === "message/rfc822" || /\.eml$/i.test(attachment.filename || "");
-      if (attached && depth < MAX_NESTED_DEPTH && Buffer2.isBuffer(content) && content.length <= budget) {
+      const attached = isAttachedMessage(attachment);
+      if (attached && Buffer2.isBuffer(content) && depth >= MAX_NESTED_DEPTH) {
+        unopened.push({ filename, reason: `nested more than ${MAX_NESTED_DEPTH} deep` });
+        continue;
+      }
+      if (attached && Buffer2.isBuffer(content) && opened >= MAX_ATTACHED_MESSAGES) {
+        unopened.push({ filename, reason: `more than ${MAX_ATTACHED_MESSAGES} attached messages` });
+        continue;
+      }
+      if (attached && Buffer2.isBuffer(content) && content.length > budget) {
+        unopened.push({ filename, reason: `past ${MAX_NESTED_BYTES / 1024 / 1024} MB of attached messages` });
+        continue;
+      }
+      if (attached && Buffer2.isBuffer(content)) {
         budget -= content.length;
+        opened++;
         let inner;
         try {
           inner = await parseWithin(content, parse, deadline);
@@ -5651,9 +5685,12 @@ ${visibleText(html.slice(0, 1e5)).text}`)) {
     hit("SEXTORTION_SUBJECT", 2, "Subject used by sextortion and account takeover scams, without a demand for payment", "SCAM");
   }
   const markup = html.slice(0, 1e5);
+  const readable = readableText(`${subject}
+${text.slice(0, 5e4)}
+${visibleText(withoutHelp(markup)).text}`);
   if (isInjection(`${subject}
 ${text.slice(0, 5e4)}
-${withoutHelp(markup)}`) || attributeText(markup).some(({ text: text2, start, end, help }) => (!help || isOrder(text2)) && isInjection(unquoted(text2), AI_ACTION_OBJECT, () => shownNear(markup, start, end)))) {
+${withoutHelp(markup)}`) || isInjection(readable) || attributeText(markup).some(({ text: text2, start, end, help }) => (!help || isOrder(text2)) && isInjection(unquoted(text2), AI_ACTION_OBJECT, () => shownNear(markup, start, end)))) {
     hit("PROMPT_INJECTION", 3, "Contains instructions addressed to an AI filter", "SPAM");
   }
   if (session.resolvedClientHostname?.endsWith(".outbound.protection.outlook.com")) {
@@ -7031,7 +7068,7 @@ function scoreResults(results, options) {
 var package_default = {
   name: "spamscanner",
   description: "Spam, phishing and malware filter for email in any language: a trainable classifier, link and attachment checks, SPF/DKIM/DMARC, and an optional AI second opinion from any LLM, including local Ollama models. Works as a library, CLI, Postfix milter or HTTP API.",
-  version: "7.0.7",
+  version: "7.0.9",
   author: "Forward Email (https://forwardemail.net)",
   bin: {
     spamscanner: "./dist/esm/cli.js"
@@ -8833,14 +8870,16 @@ var SpamScanner = class {
    * a header field ("Name: value", or an mbox "From " line) is read as the
    * body of a message with no header, so its text is the input; other
    * input mailparser cannot read gives a message with no text. Other errors
-   * are thrown.
+   * are thrown. `error` is the parser's limit the message exceeds, so that
+   * only its first parts were read (as scan() reports it, see parseError),
+   * or null.
    * @param {Buffer|string|Uint8Array|AsyncIterable} source
-   * @returns {Promise<{raw: Buffer, mail: object}>}
+   * @returns {Promise<{raw: Buffer, mail: object, error: string|null}>}
    */
   async parse(source) {
     const raw = await toBuffer(source);
-    const { mail } = await parseWithin(withHeader(raw));
-    return { raw, mail };
+    const { mail, error } = await parseWithin(withHeader(raw));
+    return { raw, mail, error: error ?? null };
   }
   /**
    * Classifier features of a parsed message.
@@ -9079,7 +9118,7 @@ var SpamScanner = class {
               message: `Attachment "${filename}" contains ${result.viruses.join(", ")}`
             });
             viruses.push(found);
-            if (attachment.contentType === "message/rfc822" || /\.eml$/i.test(attachment.filename || "")) {
+            if (isAttachedMessage(attachment)) {
               carriers.set(found, attachment.filename || "attached message");
             }
           }
@@ -9200,7 +9239,8 @@ ${visibleText(html.slice(0, config.maxLength * 2)).text}`, { maxLength: config.m
       mixed: Math.max(plainObfuscation.mixed, shownObfuscation.mixed),
       styled: plainObfuscation.styled || shownObfuscation.styled
     };
-    const attached = await expandAttachments(mail.attachments, parseErrors);
+    const unopened = [];
+    const attached = await expandAttachments(mail.attachments, parseErrors, void 0, unopened);
     const attachments = config.attachments === false ? [] : inspectAttachments(attached).filter((item) => config.macros !== false || !["macro", "pdf_active", "rtf_object"].includes(item.type)).map((item) => finding(item));
     const [links, authentication, checkedReputation, dnsbl, virusScan, toxicity, nsfw] = await Promise.all([
       this.checkLinks(checkedLinks, config, extracted.hosts, fromDomain(mail)),
@@ -9321,6 +9361,13 @@ ${visibleText(html.slice(0, config.maxLength * 2)).text}`, { maxLength: config.m
         // scanned, as {filename, reason}: the same on every try, so not a
         // failure.
         virusScanSkipped: virusScan.skipped,
+        // Attached messages not opened, past the limits attached messages
+        // are read within (nested more than three deep, more than 100
+        // attachments in them, or more than 25 MB of them), as
+        // {filename, reason}: their attachments were not inspected. Not
+        // scored (forwarded messages can be nested so), for a caller that
+        // needs a message read whole, e.g. to block what it carries.
+        unopenedAttachments: unopened,
         // Whether the message has more text than phishing.maxLength, whose
         // links were not read.
         linksTruncated,
@@ -9369,6 +9416,7 @@ export {
   scoreResults,
   segmentWords,
   spamHeaders,
-  train
+  train,
+  unwrapLink
 };
 //# sourceMappingURL=index.js.map