vLLM Under Siege: Critical DoS and SSRF Flaws Threaten Inference Stability

A cluster of high-severity vulnerabilities in vLLM exposes inference engines to SSRF, resource exhaustion, and crashes via malformed multimodal inputs. Immediate patching to v0.31.0+ is critical for platform stability.