axios CVE-2026-67314

Exploitation requires two simultaneous conditions to be met. First, the host application must already be vulnerable to a separate prototype-pollution primitive that allows an attacker to set string values on Object.prototype - this could be a distinct CVE in a dependency, a merge/assign gadget over user-supplied JSON, or similar. Second, the affected axios call site must pass an auth object that is either empty or missing username or password as own properties, using a pattern such as auth: opts.auth || {} or auth: { username: userInput } where password is omitted. Requests that explicitly supply both auth.username and auth.password as own string properties are not affected, and requests with no own auth object after mergeConfig() are also unaffected. The CVSS 4.0 AT:P metric captures this compound requirement. Exploitation is further limited by the need for network access to the service receiving the tampered outbound request, and credential disclosure specifically requires that the tampered destination be observable by the attacker.

The official CVSS 4.0 vector (AV:N/AC:L/AT:P/PR:N/UI:N/VC:N/VI:L/VA:L) scores 6.3, and the AT:P (Attack Requirements: Present) metric accurately encodes the compound-exploitation requirement. Real-world risk is lower than the score might suggest for most environments: an attacker must first achieve successful prototype pollution of the running Node.js or browser process through a distinct vulnerability before this gadget becomes exploitable. Applications that sanitize user-controlled input through JSON.parse (which cannot create functions but can create string values) and subsequently pass partial auth objects through axios are the realistic at-risk population. No CISA KEV listing exists, no public proof-of-concept has been identified, and EPSS data was not provided in the input. The primary impact is outbound request tampering - an attacker can replace or inject a Basic auth header, potentially causing downstream authentication to succeed with wrong credentials or fail entirely. Credential exfiltration is a secondary risk requiring additional application-specific conditions (e.g., the tampered request is routed to an attacker-observable endpoint and a partial real credential is present). Overall, this is a real but chained, context-dependent issue rather than a standalone critical vulnerability.

An attacker identifies a separate prototype-pollution vulnerability in a Node.js application - such as a merge or assign operation over unsanitized user input - and uses it to set Object.prototype.username and Object.prototype.password to attacker-chosen strings. When the application subsequently makes an axios request using a pattern like axios.get(url, { auth: opts.auth || {} }) where opts.auth is absent or partial, axios reads the polluted prototype values and constructs an Authorization: Basic header with the attacker's credentials, potentially bypassing or replacing the intended authentication on the downstream service call. No public proof-of-concept for this chained attack has been identified at time of analysis.