Research on network security risk detection and management protocol vulnerability mining based on fuzzing

With the increasing complexity of network protocols, traditional fuzzing methods based on random mutation are inefficient when dealing with multistate, multiconstrained protocols, resulting in limited coverage and low vulnerability trigger rates. To address this issue, we propose an improved algorithmic framework, Semantic-Aware Fuzzing (SAFuzz), which integrates semantically constrained input mutation, state-aware scheduling, dynamic feedback optimization, and vulnerability priority updating. A supporting protocol vulnerability detection and management mechanism is also designed. Experimental results show that SAFuzz significantly outperforms competing tools in both coverage and vulnerability discovery. In the HTTP/1.1 protocol, function coverage reaches 88.7%, a 20.3% improvement over American Fuzzy Lop (AFL), and path coverage increases to 71.4%. In the DNS protocol, state coverage reaches 63.7%, a 14.1 % increase over American Fuzzy Lop for Network (AFLNet) protocols. In terms of vulnerability discovery efficiency, a total of 33 HTTP/1.1 vulnerabilities were identified in a 72-hour experiment, with the average time to first vulnerability trigger reduced to 41 minutes. The vulnerability localization accuracy for the FTP protocol exceeded 81.6%, an improvement of over 20 % over AFL. By introducing semantic constraints and dynamic feedback mechanisms, SAFuzz can achieve efficient mining and risk management of protocol vulnerabilities, providing new ideas and technical paths for network security risk detection.