IMPROVING THE PERFORMANCE OF IDS USING IMPROVE FEATURE SELECTION METHOD

Abstract

Nowadays, the use of networks and especially the Internet has become a big part of daily life. According to rapid development and widespread use of network systems, diverse intrusive approaches have grown extensively in the recent years. Multiple protection techniques have been used in order to manage the security network risks. These methods do not suffice, as each of them have proven their inefficiency. Therefore, the use of intrusion detection systems as an additional defense mechanism is almost indispensable. An Intrusion Detection System (IDS) dynamically monitors the events taking place in a system, and decides whether these events are symptomatic of an attack (intrusion) or constitute a legitimate use of the system. Since the appearance of IDS multiple techniques have been proposed in order to improve the performances of these. Recently, several machine learning techniques and optimization techniques have been applied to make it efficient and to improve accuracy. In this project I am using Hybrid Binary PSO with SVM to find best subset of dataset to train our prediction model, and using this I’m improving the performance of prediction model. In this project I’m also comparing the performance of different classification algorithms like SVM, Random Forest, and Naïve baye’s.