vLLM is an inference and serving engine for large language models. From 0.11.1 until 0.32.0, the DeepSeek-OCR, DeepSeek-OCR-2, and Unlimited-OCR image path allows a remote client that can reach the OpenAI-compatible API to submit a legal image whose normalized pixels sum to zero when that image is alone in its per-step encoder batch. _parse_and_validate_image_input returns no embedding, embed_multimodal propagates that value, and sanity_check_mm_encoder_outputs triggers a fatal assertion that shuts down the engine for all users. This issue is fixed in version 0.32.0.
Weakness
The product contains an assert() or similar statement that can be triggered by an attacker, which leads to an application exit or other behavior that is more severe than necessary.
Affected Software
| Name | Vendor | Start Version | End Version |
|---|---|---|---|
| Vllm | Vllm | 0.11.1 (including) | 0.32.0 (excluding) |
Extended Description
While assertion is good for catching logic errors and reducing the chances of reaching more serious vulnerability conditions, it can still lead to a denial of service. For example, if a server handles multiple simultaneous connections, and an assert() occurs in one single connection that causes all other connections to be dropped, this is a reachable assertion that leads to a denial of service.