{"id":2291,"date":"2025-04-29T20:34:15","date_gmt":"2025-04-29T20:34:15","guid":{"rendered":"http:\/\/localhost\/?p=2291"},"modified":"2025-04-29T20:34:15","modified_gmt":"2025-04-29T20:34:15","slug":"cve-2025-46560-vllm-phi4mm-quadratic-time-complexity-in-input-token-processing-leads-to-denial-of-se","status":"publish","type":"post","link":"https:\/\/zero.redgem.net\/?p=2291","title":{"rendered":"CVE-2025-46560 vLLM phi4mm: Quadratic Time Complexity in Input Token Processing\u200b leads to denial of service"},"content":{"rendered":"
\n

Vulnerability Details<\/h2>\n
\n

Basic Information<\/h3>\n\n\n\n\n\n\n
Title<\/th>\nCVE-2025-46560 vLLM phi4mm: Quadratic Time Complexity in Input Token Processing\u200b leads to denial of service<\/td>\n<\/tr>\n
Type<\/th>\ncvelist<\/td>\n<\/tr>\n
Published<\/th>\n2025-04-30T00:24:53<\/td>\n<\/tr>\n
Last Seen<\/th>\n2025-04-30T00:58:57<\/td>\n<\/tr>\n
CVSS Score<\/th>\n6.5 (MEDIUM)<\/td>\n<\/tr>\n<\/table><\/div>\n
\n

CVSS v3 Details<\/h3>\n\n\n\n\n\n\n\n\n\n
Attack Vector<\/th>\nNETWORK<\/td>\n<\/tr>\n
Attack Complexity<\/th>\nLOW<\/td>\n<\/tr>\n
Privileges Required<\/th>\nLOW<\/td>\n<\/tr>\n
User Interaction<\/th>\nNONE<\/td>\n<\/tr>\n
Scope<\/th>\nUNCHANGED<\/td>\n<\/tr>\n
Confidentiality Impact<\/th>\nNONE<\/td>\n<\/tr>\n
Integrity Impact<\/th>\nNONE<\/td>\n<\/tr>\n
Availability Impact<\/th>\nHIGH<\/td>\n<\/tr>\n<\/table><\/div>\n
\n

CVE Information<\/h3>\n\n\n\n\n
CVE IDs<\/th>\nCVE-2025-46560<\/td>\n<\/tr>\n
CWE<\/th>\nCWE-1333<\/td>\n<\/tr>\n
Bulletin Family<\/th>\ncve<\/td>\n<\/tr>\n<\/table><\/div>\n
\n

Description<\/h3>\n
\n vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Versions starting from 0.8.0 and prior to 0.8.5 are affected by a critical performance vulnerability in the input preprocessing logic of the multimodal tokenizer. The code dynamically replaces placeholder tokens (e.g., <|audio_|>, <|image_|>) with repeated tokens based on precomputed lengths. Due to \u200b\u200binefficient list concatenation operations\u200b\u200b, the algorithm exhibits \u200b\u200bquadratic time complexity (O(n\u00b2))\u200b\u200b, allowing malicious actors to trigger resource exhaustion via specially crafted inputs. This issue has been patched in version 0.8.5.\n <\/div>\n<\/p><\/div>\n
\n

Impact Assessment<\/h3>\n\n\n\n
Base Score<\/th>\n6.5<\/td>\n<\/tr>\n
Severity<\/th>\nMEDIUM<\/td>\n<\/tr>\n<\/table><\/div>\n
\n

View full CVE details<\/a><\/p>\n<\/p><\/div>\n<\/div>\n