Design And Implementation Of Machine Learning-based Network Security Situational Awareness System For Power Enterprises

Due to the continuous complexity of information network structures within power companies,coupled with frequent security incidents,the difficulty of network security defense has significantly increased.This has led to the gradual inadequacy of existing network security measures in meeting the demands of network security protection in complex power company network environments.To assist power companies in timely identification and prediction of potential security threats,and to enhance the level of network security in the power sector,this thesis conducts in-depth research on the technology of network security situation awareness for power companies.Building upon existing research achievements,the specific work content is as follows:(1)A power enterprise network security situation evaluation model based on LIBA-DBN is proposed.In order to accurately evaluate the network security situation of power enterprises,this thesis proposes a power enterprise network security situation evaluation model based on LIBA-DBN,which optimizes the hyperparameters of DBN during training using the improved LIBA algorithm to avoid DBN getting stuck in local extremes in the solving process,effectively improving the prediction accuracy of DBN.To verify the performance of the model,experiments are conducted on the UNSW-NB15 dataset.The experimental results show that the LIBA-DBN evaluation model proposed in this thesis has improved precision and average precision in binary and multi-classification compared to some traditional machine learning models,and is better than the best of the other three models,with improvements of1.77% and 4.89%,respectively.(2)A power enterprise network security situation prediction model based on INGWO-LSTM is proposed.In order to accurately predict the network security situation of power enterprises,this thesis further studies the research results of power enterprise network security situation evaluation and proposes a network security situation prediction model based on INGWO-LSTM.The model uses an improved INGWO algorithm to optimize LSTM,effectively improving the performance of LSTM.To verify the prediction effect of the model,a series of comparative experiments are designed.The experimental results show that the proposed INGWO-LSTM prediction model has better performance in reducing training error and improving the fit degree of observed values than some traditional machine learning models.(3)Based on the above algorithm theory,this thesis finally implements a machine learning-based power enterprise network security situation awareness system.The back-end of the system is built using the Flask framework and the data storage is implemented using the My SQL database.The front-end is mainly implemented based on Vue.js.By integrating the open-source Echarts visualization library when constructing front-end components,the front-end can present the processed data information from various modules in the back-end to users in the form of charts.