Anomaly Intrusion Detection Using SVM and C4.5 Classification With an Improved Particle Swarm Optimization (I-PSO)

In the last decade, many researchers have proposed several models of classification algorithms for enhancing the accuracy performance of IDSs. However, there is a minor issue arising in the classifier's incapability to process high-dimensional data. Using several classifiers always outperforms a single classifier's performance. This paper proposes a novel intrusion detection system by classifying data with SVM as well as C4.5 decision tree algorithm. The NSL-KDD dataset is first preprocessed with principal component analysis (PCA) and later feature selected with an improved particle swarm optimization (I-PSO). This framework improved the time consumption and inaccurate feature selection issues in other methodologies. Upon simplifying features more effectively, the outcomes display an excellent agreement with the conventional PSO techniques and their results, and also produce enhanced outcomes when compared to only single classifier. The results demonstrate better performance when subject to different attack-scenarios and can be used for enterprise network security applications.