The vulnerability in vLLM (vllm-project/vllm) arises from the VideoMediaIO.load_base64() method's handling of 'video/jpeg' data URLs. The method splits the base64 data string on commas to extract individual JPEG frames but does not enforce any limit on the number of frames processed. An attacker can exploit this by submitting a single API request containing thousands of comma-separated base64-encoded JPEG frames, causing the server to decode all frames into memory simultaneously. This results in uncontrolled resource consumption, leading to an Out-of-Memory (OOM) condition and Denial of Service (DoS). The vulnerability is reachable without authentication via the OpenAI-compatible chat completions API. No official patch or remediation level is currently provided by the vendor, and no known exploits are reported in the wild. The Red Hat advisory linked does not specify a fix or mitigation status.

Successful exploitation causes a high-severity Denial of Service (DoS) by exhausting server memory resources, leading to a crash or unavailability of the affected vLLM service. The vulnerability does not impact confidentiality or integrity but results in service disruption. It is exploitable remotely without authentication, increasing the risk of attack.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Since no official fix or workaround is currently documented, users should monitor the vendor advisory for updates. Until a patch is available, consider implementing request size limits or input validation to restrict the number of frames processed in 'video/jpeg' data URLs if feasible within the deployment environment.