CVE-2025-46570
Published: 29 May 2025
Summary
CVE-2025-46570 is a low-severity Observable Timing Discrepancy (CWE-208) vulnerability in Vllm Vllm. Its CVSS base score is 2.6 (Low).
Operationally, ranked at the 39.0th percentile by exploit likelihood (below the median); it is not currently listed in the CISA KEV catalog.
This vulnerability is AI-related — categorised as NLP and Transformers; in the Adversarial Attacks risk domain.
EU & UK References
- 🇪🇺 ENISA EUVD: EUVD-2025-16189
Vulnerability details
vLLM is an inference and serving engine for large language models (LLMs). Prior to version 0.9.0, when a new prompt is processed, if the PageAttention mechanism finds a matching prefix chunk, the prefill process speeds up, which is reflected in…
more
the TTFT (Time to First Token). These timing differences caused by matching chunks are significant enough to be recognized and exploited. This issue has been patched in version 0.9.0.
- CWE(s)
AI Security AnalysisAI
- AI Category
- NLP and Transformers
- Risk Domain
- Adversarial Attacks
- OWASP Top 10 for LLMs 2025
- None mapped
- Classification Reason
- Matched keywords: llms, vllm
Related Threats
Affected Assets
Mitigating Controls
Likely Mitigating Controls AI
Per-CVE control mapping for this CVE has not run yet; the list below is derived from the weakness types (CWEs) cited in the NVD entry.
Misdirection can normalize or falsify responses to eliminate observable discrepancies that aid reconnaissance.
Observable timing discrepancies are a primary mechanism for constructing covert timing channels; analysis identifies and bounds them, limiting exploitation.
Prevents attackers from using observable differences in error responses to infer internal system details or state.