Snyk has a proof-of-concept or detailed explanation of how to exploit this vulnerability.
The probability is the direct output of the EPSS model, and conveys an overall sense of the threat of exploitation in the wild. The percentile measures the EPSS probability relative to all known EPSS scores. Note: This data is updated daily, relying on the latest available EPSS model version. Check out the EPSS documentation for more details.
In a few clicks we can analyze your entire application and see what components are vulnerable in your application, and suggest you quick fixes.
Test your applicationsLearn about Allocation of Resources Without Limits or Throttling vulnerabilities in an interactive lesson.
Start learningUpgrade vllm to version 0.31.0 or higher.
vllm is an A high-throughput and memory-efficient inference and serving engine for LLMs
Affected versions of this package are vulnerable to Allocation of Resources Without Limits or Throttling via the MiniMax M3 VL video processor in vllm/transformers_utils/processors/minimax_m3.py, which accepts request-level min_pixels supplied through mm_processor_kwargs without a numeric range check or a min_pixels <= max_pixels relationship check. The smart_resize function can enlarge every decoded frame to an attacker-controlled size, and the video-wide pixel budget (max_total_pixels) is only enforced when max_long_side_pixel is not None, which is its default value, so the budget check is skipped entirely. An API client can submit a video request with a large min_pixels value, causing disproportionate CPU and host-memory consumption during frame resizing, normalization, and tensor concatenation that may disrupt workers or the service.
Note: This is only exploitable when the server is running a MiniMax M3 VL model and the client is permitted to supply mm_processor_kwargs (i.e., --trust-request-mm-kwargs is enabled, which is the default behavior before 0.31.0).