Network Protocol Vulnerability Assessment: A Comparative Analysis of SMB Vulnerabilities in Windows Network Environments

Abstract

This study explained the development of Server Message Block (SMB) and its composition and role as a communication protocol, and delved into the SMBGhost vulnerability (CVE-2020-0796) in Windows network environments. Through strictly controlled experimental variables, it systematically quantifies the exploit success rate under different configurations and evaluates the protective effectiveness of Microsoft’s KB4551762 patch. The experimental results indicate that the exploit success rate for unpatched systems reaches 90%, while applying the patch results in a 37% decline in file transfer performance. To balance security and performance, this paper proposes a lightweight Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) hybrid detection model, which reduces the false positive rate by 63% compared to traditional Snort rules, achieves a detection accuracy of 96%, and has an inference latency of only 1.8 ms. Key findings include precise delineation of the memory corruption boundaries in the srv2.sys driver, as well as empirical validation of a worm propagation rate of 3.2 devices per second, providing an optimized solution for enterprise network defense that considers both security and performance.