Scans for LiteLLM/Starlette (CVE-2026-42271, CVE-2026-48710), (Mon, Nov 16th)
Security researchers have observed attempts to exploit CVE-2026-42271 and CVE-2026-48710 vulnerabilities in LiteLLM and Starlette through modified attack patterns.
Over the weekend, security researchers detected new exploit attempts targeting vulnerabilities in the Starlette web framework and LiteLLM. The attacks specifically target the /mcp-rest/test/tools/list endpoint, attempting to chain CVE-2026-42271 and CVE-2026-48710 to achieve remote code execution without authentication.
The attacks involve malformed Host headers containing unusual characters like '[', ']', and others to bypass access control mechanisms. These modified exploit attempts have been observed in honeypot systems, indicating active scanning for vulnerable systems.
Vulnerability Details
The core issue lies in Starlette versions prior to 1.0.1, which fail to properly validate the 'Host' header. This vulnerability, known as 'BadHost,' allows attackers to manipulate URL construction and bypass access controls. When exploited through LiteLLM, it enables unauthorized access to protected endpoints.
The problem starts with Starlette. Versions prior to 1.0.1 do not validate the 'Host' header. This vulnerability became known as 'BadHost'.
Security Recommendations
Security experts emphasize that these are not new vulnerabilities, but attackers continue to modify their techniques to evade detection and find new victims. Organizations should focus on patching systems and implementing robust Web Application Firewall (WAF) rules that flag invalid characters in hostnames rather than relying on specific exploit signatures.
Developers are advised to validate all data components before assembling them into complex entities like URLs. Reducing the attack surface through proper configuration and security hardening remains crucial as the number of vulnerabilities continues to increase.