MasihMoafi/Eyes-Wide-Shut
LLM security research: Red-teaming GPT-OSS-20B. Discovered 3 high-severity vulnerabilities including jailbreak attempts, semantic exploits, and CoT exposure.
🎯 Vulnerability: Web Security Vulnerability
💡 Defense Fix: Implement strict input validation, contextual output encoding, and enforce least-privilege authorization controls.