Anomaly based Intrusion Detection System using Particle Swarm Optimization

Abstract

In this fast-growing era of 21st century, where technology has taken over most of our aspects of life, the development of computer and networks has indeed made our life a lot easier. But this development has also brought insecurity of information on the internet and has the risk of information theft and hacking. Researchers, all over the world, are focusing on providing counter for such risks and security against hacking using different methods and techniques. There are two types of Intrusion Detection System (IDS), anomaly-based and signature-based. In the research work that is being presented here, anomaly-based IDS is modeled using Machine Learning and Swarm Intelligence algorithm to decide whether the attack is taking place or not. A methodology / algorithm based on stochastic optimization technique is proposed. It’s called Particle Swarm Optimization (PSO), which is a part of swarm intelligence. PSO is used for feature selection process after which we’ll get optimized values to decide which features are best suited for classification. For classification, K Nearest Neighbour (KNN), Support Vector Machine (SVM) and Logistic Regression (LR) classifiers were implemented. Statistical results obtained concludes that using proposed methodology KNN got an accuracy of 77%, SVM got an accuracy of 83% and LR got an accuracy of 97%.