Executive brief
OpenClaw is an AI automation platform that uses webhooks to integrate with external systems. The vulnerability allows remote attackers with network access to the webhook endpoint to gradually recover the webhook authentication token by measuring response times across many requests. This timing-based side-channel attack could lead to unauthorized access to webhooks if an attacker can measure network latency with sufficient precision. Real-world exploitation is difficult due to network jitter, but restricting webhook endpoint access and rotating tokens reduces risk.
Technical details
The vulnerability is a timing side-channel attack (CWE-208) in OpenClaw's hook token authentication. The application uses a regular (non-constant-time) string comparison to validate webhook tokens, leaking timing information based on how many characters match before the comparison fails. An attacker with network access to the hooks endpoint can send crafted requests and measure response time variations to infer the correct token byte-by-byte. This requires network reachability to the hooks endpoint and multiple probe requests; real-world latency variance makes exploitation challenging but not impossible. The fix (commit 113ebfd) replaces the vulnerable comparison with constant-time comparison and adds per-client rate-limiting (HTTP 429 responses with Retry-After headers) to throttle authentication failures. Versions 2026.2.13 and later include the patch.
Affected products
- openclaw openclaw < 2026.2.13
Timeline
- 2026-03-02: disclosed: GHSA-47q7-97xp-m272 published
- 2026-02-13: patched: Fix committed (113ebfd6a23c4beb8a575d48f7482593254506ec)
- 2026-02-15: other: GitHub advisory published by steipete
- 2026-03-05: other: NVD entry published