A Lightweight Network Intrusion Detection Model Based on Feature Selection

Network intrusion detection system (NIDS) uses all data features which contain irrelevant and redundant features. These features influence both the performance of the system and the types of attacks that NIDS detects. At the same time, they cause slow training and testing process, system resource consumption expensive as well as low true detection rate. Therefore, feature selection is an important issue in NIDS. Our research focused on mining the most useful network features for attack detection. In this paper, we proposed a new hybrid feature selection algorithm based on Chi-Square and enhanced C4.5 algorithm to build lightweight network intrusion detection system. The attributes selection technique used in the preprocessing phase to emphasize the most relevant attributes, allow making model of classification simpler and easy to understand. Verification test have been carried out by using the 1999 KDD Cup datasets. From the experiment, it is observed that significant improvement has been achieved from the viewpoint of both high true positive rate and reasonably low false positive rate while retaining low testing time.