The vulnerability in Edge0-AI Audio8_TTS arises from the ArkTtsRuntime.__init__ method within runtime.py, which resolves model paths. By manipulating the slow_decode_models argument, an attacker can supply malicious path strings that lead the runtime to resolve to files outside the intended model directory. This path traversal flaw can expose sensitive files or potentially allow an attacker to load arbitrary code, leading to remote code execution or information disclosure. With a CVSS score of 5.3, the flaw represents a moderate risk if exploited.

The issue affects the Edge0-AI Audio8_TTS product, specifically any release up to the commit reference 1d64afcc2da56222b41ddcf53de7b8bc54be1fb6. As the project follows a rolling‑release model, precise version numbers for newer builds are not disclosed, so any instance running a commit equal to or older than that should be considered vulnerable. The problem originates in the Model Path Resolution component of the onnx_runtime_0_1b_int8 module.

The CVSS score of 5.3 indicates moderate impact, while the EPSS score is not available, and the vulnerability is not listed in the CISA KEV catalog. Public proof‑of‑concept code is available, making exploitation relatively straightforward from a remote location. The vulnerability could be leveraged remotely, as the affected function is invoked during initialization and the malicious argument can be sent over the network. Given the lack of an official vendor fix, organizations need to evaluate the risk and consider mitigating controls.