The flaw resides in the TensorFlow HDF5 library’s handling of plugin loading. An attacker who can run code with limited privileges can manipulate the search path for plugins, causing the library to load an attacker‑controlled plugin from an unsecured location. This results in execution of arbitrary code with the privileges of the target user, enabling local privilege escalation.

This vulnerability affects all installations of TensorFlow that use the HDF5 library to load plugins. No specific version ranges are supplied, so any TensorFlow release prior to the security fix referenced in the commit developers should be considered vulnerable. The vendor product is TensorFlow.

The CVSS base score of 7.8 indicates a high severity issue. The EPSS score is reported as < 1 %, suggesting a very low probability of exploitation in the wild at this time. The vulnerability is not listed in the CISA KEV catalog, further indicating it is not a known exploited flaw. Exploitation requires an attacker to already have the ability to run code with low privileges on the host; from that position, manipulating the plugin search path permits escalation to higher privilege levels. The local nature of the attack limits the attack surface to machines where the user has some access, and the exploit chain does not rely on network exposure.