Red Hat Ai Inference Server
Monthly
Privilege escalation in Red Hat Data Science Pipelines (DSP) allows an authenticated namespace editor to gain node-root access by submitting a crafted Argo Workflow via the V1 API path. The API server, acting as a confused deputy (CWE-266), creates pods with elevated privileges on the attacker's behalf, effectively breaking out of Kubernetes namespace isolation. Successful exploitation enables arbitrary code execution and full control over the underlying cluster node; no public exploit has been identified at time of analysis, and this is not listed in CISA KEV.
Image input manipulation in vLLM's multimodal preprocessing pipeline allows remote, unauthenticated network attackers to craft images with specific EXIF orientation or PNG tRNS transparency metadata that, when converted to RGB by vLLM, produces semantically altered image content fed to the LLM - affecting the integrity of inference outputs and potentially the reliability of the inference service. Affected deployments include Red Hat AI Inference Server across RHEL AI 3 and Red Hat OpenShift AI (RHOAI) environments. No public exploit code has been identified at time of analysis and the vulnerability is not listed in the CISA KEV catalog; however, sensitive inference workloads processing user-supplied images (e.g., document classification, content moderation) face a higher practical risk from subtle input distortion attacks.
A remote attacker can trigger a heap out-of-bounds write in FFmpeg's DVD subtitle parser by providing a crafted MPEG-PS/VOB file containing a malicious subtitle stream. This signed integer overflow flaw leads to application crash or potential arbitrary code execution, affecting any software that relies on FFmpeg for media parsing. No public exploit is known, and EPSS indicates a low exploitation probability (0.04%).
Privilege escalation in Red Hat Data Science Pipelines (DSP) allows an authenticated namespace editor to gain node-root access by submitting a crafted Argo Workflow via the V1 API path. The API server, acting as a confused deputy (CWE-266), creates pods with elevated privileges on the attacker's behalf, effectively breaking out of Kubernetes namespace isolation. Successful exploitation enables arbitrary code execution and full control over the underlying cluster node; no public exploit has been identified at time of analysis, and this is not listed in CISA KEV.
Image input manipulation in vLLM's multimodal preprocessing pipeline allows remote, unauthenticated network attackers to craft images with specific EXIF orientation or PNG tRNS transparency metadata that, when converted to RGB by vLLM, produces semantically altered image content fed to the LLM - affecting the integrity of inference outputs and potentially the reliability of the inference service. Affected deployments include Red Hat AI Inference Server across RHEL AI 3 and Red Hat OpenShift AI (RHOAI) environments. No public exploit code has been identified at time of analysis and the vulnerability is not listed in the CISA KEV catalog; however, sensitive inference workloads processing user-supplied images (e.g., document classification, content moderation) face a higher practical risk from subtle input distortion attacks.
A remote attacker can trigger a heap out-of-bounds write in FFmpeg's DVD subtitle parser by providing a crafted MPEG-PS/VOB file containing a malicious subtitle stream. This signed integer overflow flaw leads to application crash or potential arbitrary code execution, affecting any software that relies on FFmpeg for media parsing. No public exploit is known, and EPSS indicates a low exploitation probability (0.04%).