Improved Cyber Attack Detection Using MLB-FQS: A Novel Modified Lagrange Butterfly-based Fuzzy Quasi-Linear SVM Algorithm

Abstract

With the rapid enhancement of communication and computer networks, internet-based technologies provide numerous convenient and accessible services to the public all over the world. Meanwhile, the types and number of cyber attacks have been increasing day by day and create a serious threat to secure most confidential data. Hence, securing data is considered a crucial task and nowadays intrusion detection system is one of the most effective and promising techniques to provide secure data by detecting malicious data. However, it becomes extremely hard in preventing the network system from attack due to imbalanced data that results in low detection and poor accuracy rates. Therefore, a novel intrusion detection mechanism is proposed to accurately detect cyber intrusion with high detection accuracy rate. The Intrusion detection mechanism comprises three significant phases namely the data pre-processing phase, the feature selection phase as well as attack detection phase. Various pre-processing techniques namely data normalization, data conversion as well as one-hot encoding process are utilized to provide an appropriate format that is suitable for analysis. In the feature selection phase, a Modified Lagrange Butterfly (MLB) optimization is employed to select optimal features for effective detection. Finally, a Fuzzy Quasi-Linear Support Vector Machine (FQL-SVM) classifier is employed to determine the network as two different types of classes namely the abnormal or malicious classes. After examining on malware dataset, the experimental analysis demonstrates a high classification accuracy rate of 98.41% which is better than other comparative techniques.