A vulnerability in the MaxKB AI assistant allows authenticated attackers to bypass process execution restrictions, potentially leading to unauthorized system command execution.

The application utilizes an LD_PRELOAD sandbox to restrict subprocess creation, but it fails to hook the fexecve system call. An authenticated attacker capable of executing tool code can leverage this oversight to launch processes outside the intended security policy.

The ability to execute unauthorized processes represents a significant security risk to the host environment. While the CVSS score of 6.3 identifies this as a medium severity issue, the potential for sandbox escape could allow an attacker to pivot from the AI assistant to the underlying host, resulting in data exfiltration or system compromise.

Immediate Action: Update 1Panel-dev MaxKB to version 2.10.6-lts or later to implement the necessary security hooks.

Proactive Monitoring: Review application logs for unusual system calls or unexpected process execution patterns that deviate from standard AI assistant behavior.

Compensating Controls: Ensure the application runs with the least privilege necessary, utilizing containerization or restricted service accounts to limit the impact of a potential sandbox escape.

Public Exploit Available: Unknown

Organizations utilizing MaxKB must prioritize the update to version 2.10.6-lts to ensure that the sandbox environment is correctly configured. Failure to patch allows authenticated users to potentially bypass critical security controls, significantly increasing the risk of unauthorized system-level operations.