CVE-2026-73560
MediumCVSS 6.5Exploitation Probability (EPSS)
Low risk25th percentile - higher than 25% of all known CVEs
Summary
vLLM prior to 0.26.0 in the MiMoV2OmniMultiModalProcessor passes attacker-controlled image and audio strings through _fetch_image, requests.get, and Image.open instead of MediaConnector, bypassing allowed_media_domains and allowed_local_media_path protections. This allows server-side requests and reads of arbitrary files accessible to the vLLM process.
Risk Assessment
Risk of unauthorized file access and server-side requests, potentially leading to data leakage or attacks on internal resources.
Recommendation
Upgrade vLLM to version 0.26.0 or later.
Original NVD description (English source)
vLLM is an inference and serving engine for large language models. Prior to 0.26.0, the MiMoV2OmniMultiModalProcessor in vllm/transformers_utils/processors/mimo_v2_omni.py passes attacker-controlled image and audio strings through _fetch_image, requests.get, and Image.open instead of MediaConnector, bypassing allowed_media_domains and allowed_local_media_path protections and allowing server-side requests and reads of arbitrary files accessible to the vLLM process. This issue is fixed in version 0.26.0.