An Efficient Feature Selection Method for Network Attack Detection Using PSO-Based Wrapper Technique

Abstract

Given that internet usage and connectivity are in such great demand right now, the steep increase in network attacks has been a big source of concern for cyber security. Fog computing can provide low-latency services for cloud and mobile users as an add-on to cloud computing. Fog devices may experience security difficulties due to the proximity of the end users to the fog nodes and the lack of suitable computing capacity. Fog computing system destruction may result from conventional network threats. Applying Intrusion Detection Systems (IDS) directly to the fog computing platform may be inappropriate given the vast research on their use in conventional networks. It is crucial to construct an intrusion detection system model over huge datasets in the fog computing environment because nodes of the fog frequently generate enormous amounts of data. An intrusion detection system (IDS), a strategic intrusion prevention innovation that can be used in the fog computing platform and utilize machine learning techniques for network anomaly detection and network event classification threat, has shown to be effective and efficient in defending against some of these network attacks. In order to reduce time complexity and create an improved model that can predict outcomes more accurately, this paper presented a Particle Swarm Optimization (PSO) Algorithm Wrapper-Based feature selection and Nave Bayes for Anomaly Detection Model in a Fog Environment. It uses the Security Laboratory knowledge Discovery Dataset (NSL-KDD). According to the data, the developed system performs better overall, with an accuracy rate of 98.27 percent and a false positive rate of just 1.6%. The results demonstrate that the suggested strategy outperforms comparable approaches in the literature.