ACON is a widely-used library of tools for machine learning that focuses on adaptive correlation optimization. A potential vulnerability has been identified in the input validation process, which could lead to arbitrary code execution if exploited. This issue could allow an attacker to submit malicious input data, bypassing input validation, resulting in remote code execution in certain machine learning applications using the ACON library. All users utilizing ACON’s input-handling functions are potentially at risk. Specifically, machine learning models or applications that ingest user-generated data without proper sanitization are the most vulnerable. Users running ACON on production servers are at heightened risk, as the vulnerability could be exploited remotely. As of time of publication, it is unclear whether a fix is available.

Multiple stack-based buffer overflows in (a) acon.c, (b) menu.c, and (c) child.c in Acon 1.0.5-5 through 1.0.5-7 allow local users to execute arbitrary code via (1) a long HOME environment variable or (2) a large number of terminal columns.