Increasing Feature Selection Accuracy through Recursive Method in Intrusion Detection System

Abstract

Artificial intelligence semi supervised-based network intrusion detection system detects and identifies various types of attacks on network data using several steps, such as: data preprocessing, feature extraction, and classification. In this detection, the feature extraction is used for identifying features of attacks from the data; meanwhile the classification is applied for determining the type of attacks. Increasing the network data directly causes slow response time and low accuracy of the IDS. This research studies the implementation of wrapped-based and several classification algorithms to shorten the time of detection and increase accuracy. The wrapper is expected to select the best features of attacks in order to shorten the detection time while increasing the accuracy of detection. In line with this goal, this research also studies the effect of parameters used in the classification algorithms of the IDS. The experiment results show that wrapper is 81.275%. The result is higher than the method without wrapping which is 46.027%.