Research On Intrusion Detection Method Based On Joint Symmetric Uncertainty And Hyperparameter Optimization

An intrusion detection system(IDS)identifies network intrusions by monitoring massive amounts of network data for abnormal traffic,ensuring network security.However,IDS data faces the problem of high dimensionality,and the feature differences in different network systems and attack environments lead to poor intrusion detection performance and low algorithm portability.To address this issue,the intrusion detection method based on joint symmetric uncertainty and hyperparameter optimization fusion neural network is studied in this paper.Firstly,a joint symmetric uncertainty feature selection algorithm based on the idea of symmetric uncertainty and Markov blanket is proposed,which fully considers the correlation between data features and class labels,redundancy between features,and correlation between feature combinations and class labels.This extracts the optimal feature subset to solve the problem of high-dimensional data.Secondly,a CNN-BiLSTM model that fuses convolutional neural network(CNN)and bidirectional long short-term memory neural network(Bi LSTM)is used as the classifier.This model combines the ability of CNN to extract data spatial features through convolutional layers and the ability of LSTM to extract data temporal features through gate functions,thereby improving the classifier’s classification performance.Additionally,the introduction of Bi LSTM with the ability to learn bidirectional time series information further enhances the performance of the classification model.Finally,by improving the particle swarm optimization(PSO)algorithm and applying it to the automatic optimization of classifier hyperparameters,the generalization ability of the classifier is greatly improved,making it suitable for different forms of intrusion detection datasets under different acquisition environments,and ensuring the best classification performance.The proposed algorithm is validated on the KDD99 dataset and the UNSWNB15 dataset,and its effectiveness and superiority are verified through multiple evaluation metrics.