Semantic-Aware Fuzzing: an Empirical Framework for LLM-Guided, Reasoning-Driven Input Mutation

Security vulnerabilities in Internet-of-Things devices, mobile platforms, and autonomous systems remain critical. Traditional mutation-based fuzzers -- while effectively explore code paths -- primarily perform byte- or bit-level edits without semantic reasoning. Coverage-guided tools such as AFL+...