An arbitrary code execution vulnerability in SpeechBrain allows unauthenticated attackers to execute malicious code by providing a crafted CKPT.yaml file.

The application utilizes PyYAML with an unsafe loader to parse checkpoint metadata files. An attacker can inject malicious Python object construction tags into a CKPT.yaml file, which triggers code execution during the candidate enumeration process in Checkpointer.recover_if_possible().

This vulnerability allows for full remote code execution on systems processing untrusted checkpoint files. Given the CVSS score of 8.8, the potential for complete system compromise is high. This poses a severe risk to data integrity, system availability, and the confidentiality of sensitive machine learning models or infrastructure.

Immediate Action: Update the SpeechBrain package to version 1.1.1 or later to implement secure YAML parsing practices.

Proactive Monitoring: Monitor filesystem activity for unexpected creation or modification of CKPT.yaml files in checkpoint directories.

Compensating Controls: Ensure that directories used for storing and loading checkpoints have strict access controls and are not writable by unauthorized users or processes.

Public Exploit Available: No (exploit_available: false)

The severity of this vulnerability necessitates an immediate update to version 1.1.1. Organizations utilizing SpeechBrain must prioritize patching to prevent potential remote code execution, as the current implementation does not adequately validate the integrity of checkpoint metadata files.