Red Hat Security Advisory: Red Hat AI Inference 3.4.4 (spyre)

CVE-2026-34753 is a server-side request forgery (SSRF) vulnerability in the vLLM batch runner component of Red Hat AI Inference Server 3.4.4 (rocm). An attacker with the ability to control batch input JSON can cause the server to make arbitrary HTTP or HTTPS requests to internal services, potentially leading to information disclosure or further system compromise. The vulnerability is identified by CWE-918 and related CWEs indicating issues such as improper access control and memory corruption. Red Hat has classified the severity as high but currently does not provide a patch or mitigation that meets their criteria for ease of use, applicability, and stability. The advisory notes that the vulnerability allows bypassing access controls by making the server act as a proxy to internal resources.

The SSRF vulnerability allows attackers to coerce the server into making arbitrary HTTP/HTTPS requests, potentially exposing internal services and sensitive information such as cloud metadata or internal APIs. This can lead to information disclosure and may facilitate further compromise of the host system. The impact includes confidentiality loss and limited availability impact. No known exploits are reported in the wild at this time.

Red Hat currently does not offer an official fix or mitigation that meets their criteria for this vulnerability. Users are advised to monitor Red Hat advisories for updates. Until a fix is available, restricting access to the vulnerable service and limiting input sources may reduce risk, but no vendor-approved mitigation is currently provided.