ESA AnomalyMatch prior to version 1.3.1 loads model checkpoint files from session directories using the Python function torch.load without any restrictions on the payload, allowing malicious data to be deserialized and executed as arbitrary code. The vulnerability is a classic unsafe deserialization flaw that can be abused to compromise confidentiality, integrity, and availability of the system by running attacker supplied code in the host environment.

The affected product is ESA’s AnomalyMatch software, specifically versions earlier than 1.3.1. No other vendors or product versions are listed in the CNA data.

The EPSS score is not available and the vulnerability is not listed in CISA’s KEV catalog, giving no clear indications of current exploitation activity. No CVSS score is provided. Based on the description, the likely attack vector involves an adversary that has the ability to place a crafted checkpoint file in a session directory or upload it through an existing API that accepts model files; this would trigger the unsafe torch.load call and allow code execution. The vulnerability is classed as a remote code execution risk with potentially high severity if exploitation conditions are met.