Security-by-Design for LLM-Based Code Generation: Leveraging Internal Representations for Concept-Driven Steering Mechanisms

Security-by-Design for LLM-Based Code Generation: Leveraging Internal Representations for Concept-Driven Steering Mechanisms explores A mechanism to enhance security in LLM-based code generation by steering internal representations towards secure outputs.. Commercial viability score: 8/10 in Code Generation Security.