- 期刊
Research on Abnormal Traffic Intrusion Detection for Power Generation Enterprise Network
摘要
The normal functioning of the power generation enterprise network is closely tied to the functioning of the power generation system, and the necessity of detecting abnormal traffic for intrusion prevention cannot be overstated. Firstly, aiming at the imbalance between abnormal and normal traffic in the power generation enterprise network, this paper proposed an improved synthetic minority oversampling technique (ISMOTE) method to process the balance of the dataset. Then, an improved white shark optimizer (WSO) was designed to optimize the extreme learning machine (ELM) parameters, i.e., using the improved white shark optimizer-extreme learning machine (IWSO-ELM) algorithm to realize abnormal traffic intrusion detection. It was found that the dataset, after applying the ISMOTE, achieved better performance in intrusion detection. The IWSO algorithm improved the intrusion detection effect of ELM by more than 10%. The F1 value of the IWSO-ELM algorithm on the NSLKDD and UNSWNB15 datasets reached 95.10% and 99.34%, respectively, which was better than the decision tree, support vector machine, and other methods. The findings prove the effectiveness of the IWSO-ELM method, making it applicable to implementation in real-world power generation enterprise networks.