Threats Tagged 'cve-2026-73556'

Red Hat AI Inference

Red Hat AI Inference 3.4.5 (rocm) includes multiple security vulnerabilities, including CVE-2026-34993, which affects the AIOHTTP Python framework. This vulnerability allows potential arbitrary code execution when untrusted input is loaded via the CookieJar.load() function. Exploitation requires an application to be configured to load attacker-controlled files, which is not typical in default Red Hat deployments. No official fixes are currently available for these vulnerabilities in Red Hat AI Inference 3.4.5 (rocm).

Red Hat AI Inference 3.4.5 (cuda) is referenced in a security advisory listing multiple CVEs including CVE-2026-5241. The advisory does not provide specific details or fixes for CVE-2026-5241 itself. One related vulnerability, CVE-2026-34993 in the AIOHTTP Python framework, allows potential arbitrary code execution via untrusted input to the CookieJar.load() function, but exploitation requires specific application configurations. No official patch or fix is currently provided for Red Hat AI Inference 3.4.5 (cuda) regarding these vulnerabilities.

A security advisory for Red Hat AI Inference 3.4.5 (cpu) addresses multiple vulnerabilities including CVE-2026-34993, which involves a flaw in the AIOHTTP Python asynchronous HTTP framework. This flaw allows potential arbitrary code execution when untrusted input is loaded via the CookieJar.load() function. Exploitation requires an application to be configured to process attacker-controlled files, which is not common in default Red Hat deployments. The advisory does not list any fixes currently available for these vulnerabilities.

CVE-2026-73556 is a denial-of-service vulnerability in the vllm project's lm-format-enforcer backend. The backend compiles attacker-supplied regular expressions without a timeout or buildability check, allowing a crafted regex to stall the engine worker by consuming excessive CPU. This issue affects versions of vllm prior to 0.26.0 when the lm-format-enforcer backend is explicitly enabled. The default backend is not affected. The vulnerability results in unauthenticated denial of service by hanging the structured-output compile step and blocking concurrent requests.