CVE-2025-46570
LOWDescription
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 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.
Is your site exposed to CVE-2025-46570?
Run a free security scan — no signup, results in seconds.
CVSS v3.1 Score
Weakness Type (CWE)
Affected Products
| Vendor | Product |
|---|---|
| vllm | vllm |
References
Frequently Asked Questions
What is CVE-2025-46570? +
How severe is CVE-2025-46570? +
What products are affected by CVE-2025-46570? +
How do I check if I'm vulnerable to CVE-2025-46570? +
Related Vulnerabilities
When verifying a mac with a ChunkedMacVerification object, Tink compares the resulting tag with non constant time comparison. This potentially …
An observable timing discrepancy in the ASP could allow a privileged attacker to perform a brute-force attack against the hash …
An observable timing discrepancy in the ASP could allow a privileged attacker to perform a brute-force attack against the hash …
cryptography is a package designed to expose cryptographic primitives and recipes to Python developers. From 44.0.0 until 50.0.0, pkcs7_decrypt_der, pkcs7_decrypt_pem, …
PyQuorum is a cryptographic library for secret sharing and key management. Prior to 0.2.1, the mul_mod function implements multiplication via …
In Bouncy Castle for Java before 1.86, HQC leaked secret-derived data through two side channels: its GF(2^8) arithmetic used lookup …
