False Data Injection Attacks Intrusion Simulation Applying Semi-Supervised for Power System Based on Knowledge Graph

With the increased risk of power systems facing external attacks, the all-round simulation of conventional attack patterns has become an important demand for current power system security protection against conventional cyber attacks. This paper proposes a false data injection attack (FDIA) simulation method for power systems, which is safer, faster, more comprehensive and more accurate than the traditional FDIA monitoring and defense methods. The proposed method requires firstly to simulate the key parameters of the actual power system equipment, the actual operation process, the equipment interactions, the simulation and the actual equipment operation status to maintain consistency. Then, the simulated system is subjected to the "black-box attack" and "white-box attack" defined in this paper. After that, we obtain the key equipment parameters, state changes, and system alarms of the system after the attack. Finally, two semi-supervised learning algorithms XGboost and support vector machine(SVM) are required to improve the accuracy of this attack method. The advantage of the proposed method in this paper is that the attack can be identified by tracing the root cause in reverse, and furthermore, mutual verification can be done by the original initiated attack, and finally, the all-round analysis of the attack and the impact of the intrusion can be accomplished from the attacker’s point of view by using the knowledge graph visualization.