LLM Security and Safety: Insights from Homotopy-Inspired Prompt Obfuscation

Luis Lazo, Hamed Jelodar, Roozbeh Razavi-Far

In this study, we propose a homotopy-inspired prompt obfuscation framework to enhance understanding of security and safety vulnerabilities in Large Language Models (LLMs). By syste…