Optimizing Network Security: Feature Selection Meets Quantum SVM

IDSs are essential to network security and help detect malicious activities in today's computer networks. However, conventional machine learning IDSs are sensitive to degradation of performance in high dimensional data, high data complexity and high data redundancy. To overcome these difficulties, this paper aims at suggesting a hybrid ID framework using different FS techniques and a Quantum Support Vector Machine (QSVM) for the binary attack classification. The proposed framework is evaluated by testing it using UNSW-NB15. Preprocessing techniques like hierarchical encoding, missing value imputation, and min-max normalization are used during preprocessing to enhance data quality and consistency. To find the most informative network traffic features, three feature selection methods are used: correlation analysis, chi-square test and recursive feature elimination (RFE). The accuracy of the classification of the QSVM is compared to supervised machine learning classifiers. The results of the experiments indicate that the proposed QSVM method gives the best performance over all the feature selection methods and also with respect to traditional classifiers. When it comes to classification performance, RFE with QSVM achieves the best results (99.67% accuracy, 99.61% recall, 99.73% specificity, 99.69% precision, 99.65% F1 and 0.27% false positive rate). The results of the proposed hybrid QSVM solution are presented to be effective and reliable in next generation network intrusion detection system. This is because quantum support vector machine (SVM) is capable of handling high-dimensional data.This is because quantum support vector machine (SVM) can process high dimensional data.