LLM Embedding-Based Attribution (LEA): Quantifying Source Contributions to Generative Model'S Response for Vulnerability Analysis

Security vulnerabilities are rapidly increasing in frequency and complexity, creating a shifting threat landscape that challenges cybersecurity defenses. Large Language Models LLMs have been widely adopted for cybersecurity threat analysis. When querying LLMs, dealing with new, unseen...