Abstract
Due to the increasing complexity and diversification of cyberattack methods, traditional security defense systems can no longer meet the global demand for dynamic risk perception. Aiming at the problem of lagging and inaccurate risk assessment of situational awareness, this paper proposes a graph-based network security situational awareness and risk assessment method. In the research process based on knowledge graph construction and graph cut-in learning, the effective fusion and reasoning of multi-dimensional security information are realized. The experimental results show that different methods have different identification events for database, server and network information, but the knowledge graph-driven method has a wider recognition range, with an abnormal signal recognition ratio of 84.12~89.36% and a risk assessment accuracy of 90.72~92.16%, which is better than that of normal information network identification method of 72.36~82.11% and 80.11~84.19%. Under the recognition driven by knowledge graph, the improvement degree and situational awareness of the entire network security situation were improved by 90.12~92.06% and 86.03~89.16, respectively, which were significantly better than the results of the normal information network recognition method (82.05~87.02% and 81.02~85.12). Therefore, the knowledge graph-driven method can improve the level of network security situational awareness, accurately identify security risks, and provide support for computer network prevention.