An efficient Network Anomaly detection based on PSO-Based Wrapper Feature Selection method and Bagging Technique

Abstract

   The current high demand for internet usage has led to an increased rate of attacks in different networks, which is a major concern for cybersecurity. To complement cloud computing, fog-computing offers low-latency services to cloud and moving users. However, fog devices might face security-related challenges regarding their adjacency to end users and insufficient computational power. Moreover, most common network threats may lead to the compromise of fog computing systems. Despite extensive research on applying intrusion detection systems (IDS) in traditional networks, it may not be appropriate to immediately apply them to the fog-computing. Developing efficient intrusion detection system that can handle massive databases is important in fog computing, as fog nodes generate large volumes of data. To combat network attacks, intrusion detection systems (IDS) could be deployed in fog computing, which use (ML) machine learning methods to detect network anomalies and classify threat events, proving to be effective and efficient. The present research presents a novel approach based on Particle Swarm Optimization (PSO) and Wrapper-Based feature selection and also Bagging technique for detecting intrusion in a fog-environment, using the Knowledge Discovery Dataset from Security Laboratory (NSL-KDD). This approach reduces time complexity and produces a more accurate model for outcome prediction. The outcomes demonstrate that the presented methodology outperforms related methods from related literature, achieving an overall 98.99% accuracy and 1.5% (FP) false positive rate.