Launch Week Day 1: Announcing Security Design Review
MEDIUM 4.2 npm

fast-jwt has a ReDoS when using RegExp in allowed* leading to CPU exhaustion during token verification

GHSA-cjw9-ghj4-fwxf · CVE-2026-35041

Published · Modified

Description

⚠️ IMPORTANT CLARIFICATIONS

Affected Configurations

This vulnerability ONLY affects applications that:

  • Use RegExp objects (not strings) in the allowedAud, allowedIss, allowedSub, allowedJti, or allowedNonce options
  • Configure patterns susceptible to catastrophic backtracking
  • Example: allowedAud: /^(a+)+X$/ ← VULNERABLE
  • Example: allowedAud: "api.company.com" ← SAFE

Not Affected

  • Applications using string patterns for audience validation (most common)
  • Applications using safe RegExp patterns without nested quantifiers
  • Default fast-jwt configurations

Assessment Guide

To determine if you're affected:

  1. Check ifallowedAud, allowedIss, allowedSub, allowedJti, or allowedNonce use RegExp objects (/pattern/ or new RegExp())
  2. If yes, review the pattern for nested quantifiers like (a+)+, (.*)*, etc.
  3. If no RegExp usage, you are NOT affected

Summary

A denial-of-service condition exists in fast-jwt when the allowedAud verification option is configured using a regular expression.

Because the aud claim is attacker-controlled and the library evaluates it against the supplied RegExp, a crafted JWT can trigger catastrophic backtracking in the JavaScript regex engine, resulting in significant CPU consumption during verification.

This occurs with a validly signed JWT, making the issue exploitable in authenticated contexts such as:

  • API gateways

  • authentication middleware

  • service-to-service communication

  • OAuth / OIDC token validation pipelines


Affected Component

  • Library: fast-jwt

  • Version tested: 6.1.0

  • Runtime: Node.js v24.13.1

  • Feature: claim validation using allowedAud: RegExp


Impact

CPU exhaustion / Denial of Service

A crafted JWT causes verification to take multiple seconds per request due to catastrophic regex backtracking.

Measured verification times

Input size (n) | Verification time -- | -- 24 | ~123 ms 28 | ~1.97 s 30 | ~7.85 s

This is sufficient to:

  • block Node.js event loop threads

  • degrade API throughput

  • cause cascading service failures

  • increase serverless execution costs

  • saturate authentication infrastructure


Root Cause

The library allows regular expressions in claim validation:

allowedAud: /^(a+)+X$/

The aud claim is attacker-controlled:

aud = "a".repeat(n) + "Y"

This creates catastrophic backtracking in the JavaScript regex engine.

Verification time grows exponentially as input length increases.


Exploitability

Attack requires:

  • a valid signed JWT (post-authentication context)

  • attacker control over the aud claim

  • a vulnerable regex configured by the application

Common real-world scenarios

  • shared HS secrets

  • internal JWT issuance

  • microservice authentication

  • OAuth / OIDC custom audiences

  • internal service tokens


Proof of Concept

Reproduction steps

  1. Install fast-jwt

  2. Configure verifier with a RegExp in allowedAud

  3. Send a valid signed JWT with adversarial aud

Attached artifacts

  • poc-suite-redos-fastjwt.js

  • evidence-redos-fastjwt.json

Observed behavior

Verification CPU time increases from milliseconds to multiple seconds as input length grows.


Security Classification

  • CWE-1333: Inefficient Regular Expression

  • CWE-400: Uncontrolled Resource Consumption

  • Class: Authenticated Denial of Service


Expected Behavior

The library should prevent unbounded CPU work on attacker-controlled claims.

Possible mitigations

  • safe-regex validation

  • maximum length enforcement for claims

  • regex complexity limits

  • documentation warning about ReDoS risks when using RegExp-based validation


Notes

Signature verification occurs before claim validation, therefore this is not a pre-authentication DoS.

However, the vulnerability remains exploitable in authenticated or token-bearing contexts and can significantly impact production environments.

Ready to move

Start Securing

Free, no credit card | First findings in minutes