Intrusion Detection Using Enhanced Transductive Support Vector Machine

Abstract

The world is getting more interconnected and reliant on the Internet and the services it provides today. The protection of networks and apps from unauthorized attacks is one of the biggest difficulties in internet communication. Numerous solutions have been put out to deal with security concerns, yet the vast majority of these solutions consistently fall short of rapidly and effectively detecting security threats. In order to detect new attacks with high accuracy, a method for intrusion detection employing machine learning techniques is proposed in this article. Here, the Enhanced Transductive Support Vector Machine (ETSVM) method is used to classify the data in order to more accurately detect the different types of intrusion attacks. The more pertinent and ideal features are chosen using the Improved Glowworm Swarm Optimization (IGSO) technique. This method performs better at detecting intrusions on the KDD CUP99 and CSE-CIC-IDS2018 datasets. Precision, recall, and accuracy are used to assess the proposed model's performance in identifying the four types of cyber attacks-DoS, U2R, R2L, and Probe. In order to validate the proposed methodology, comparative findings are presented.