KAS-IDS: A Machine Learning based Intrusion Detection System

The Internet has become an integral part of our life as we perform day-to-day work like e-Banking, e-Education and e-Commerce. With this, the threat of attackers and hackers has also been increasing. A crucial part has been played by an Intrusion Detection System (IDS) detect such malicious acts. Unfortunately, most of the commercial IDSs are based on misuse based that are developed to capture the already known attacks only. These require frequent updation of signatures and have minimum capacity to capture new attacks. Therefore, anomaly-based IDS is an effective alternative for this problem. Many of the researchers adopt various techniques to enhance the efficiency of the IDS. However, the false alarm rate and detection rate are challenging issues.This paper proposes a technique called KAS-IDS, i.e. K-Means and Adaptive SVM based Intrusion Detection System. In the first step, the clusters of data have been made using K-Means and second, the classification has been performed using adaptive SVM. A well-known dataset has been used to perform the experiments. The outcomes represents that our approach has better performance as compared to the individual algorithm when it comes to detection accuracy and false alarm rate.