vLLM: Harmony tool continuations drop `cache_salt` — restoring a cross-tenant prefix-cache membership oracle
CVE-2026-105752 · GHSA-935w-9g4m-p28p · published 2026-10-06
- Severity
- Low
- CVSS score
- 3.1
- Chance of exploitation (EPSS)
- 0.22%
- Fixed in
- 0.30.0
- Exploitation No known exploitation
- Automatable No
- Technical impact Partial
- CISA SSVC
An authenticated tenant of a shared deployment can recover whether a guessed post-tool prompt or history was processed by another tenant, with exact cached-token counts rather than noisy latency — the exact prompt-membership oracle cachesalt is documented to prevent. It defeats the multi-user prefix-cache isolation guarantee for salted Harmony tool sessions. Preconditions: a GPT-OSS Harmony model on /v1/responses; prefix caching enabled (default); an operator-enabled built-in or MCP tool server; the victim sets cachesalt and triggers at least one tool continuation; and the attacker can reconstruct the post-tool history closely enough to match the token prefix.
Upgrade vllm to 0.30.0 or later. If you start it from an MCP config with npx or uvx, pin that version so an old cached copy is not used.
Propagate request.cache_salt into every Harmony (and Parsable) tool-continuation re-submission — at the continuation call site call , mirroring the correct turn-1 call. Carry the salt on the HarmonyContext (thread the originating request into the context) so no continuation path can omit it:
| Package | Registry | Affected | Fixed in |
|---|---|---|---|
| vllm | PyPI | < 0.30.0 | 0.30.0 |
- CriticalPhoenix: Format String Injection in applyChatTemplatearize-phoenix
- CriticalFlowise: Unauthenticated OAuth2 token refresh endpoint returns access tokens — enables token theft for any connected serviceflowise
- CriticalMLflow: Environment variable injection in AI Gateway secrets enables server-side credential exfiltrationmlflow
- HighConfig revision hashes could expose password verifiersopenclaw
- Highn8n: Shared Credential Header Leak via HTTP Request Pagination Expressionn8n, n8n, n8n
- HighWorkspace dotenv files could override provider credentialsopenclaw
- CriticalvLLM: OpenAI auth bypassvllm
- CriticalvLLM has RCE In Video Processingvllm
- CriticalvLLM Allows Remote Code Execution via PyNcclPipe Communication Servicevllm
- CriticalvLLM Vulnerable to Remote Code Execution via Mooncake Integrationvllm
- CriticalCVE-2025-24357 Malicious model remote code execution fix bypass with PyTorch < 2.6.0vllm
- CriticalvLLM Deserialization of Untrusted Data vulnerabilityvllm
The impact, patch and workaround text comes from the advisory, shortened (GitHub Advisory Database, CC BY 4.0). Proof-of-concept and reproduction sections and code are left out; the original advisory has them.