RETRACTED: Enhancing Multi-Class Intrusion Detection through Hybrid Auto Encoder-Deep Neural Network Classifiers: A Comprehensive Analysis of Class Imbalance Mitigation Strategies Using Data Resampling Techniques

Abstract

Network and cloud environments must be fortified against a dynamic array of threats, and intrusion detection systems (IDSs) are critical tools for identifying and foiling hostile efforts. Systems for detecting intrusions, classified as anomaly-based or signature-based, have added deep learning models to their repertoire. A significant change has occurred in these systems recently. For the popular anomaly-based intrusion detection system (IDS), which leverages machine learning, attack detection accuracy has shown to be outstanding. We demonstrate with our proposed method that deep learning has enabled unprecedented success in identifying known and unknown threats within cloud systems. However, the intrusion detection benchmark datasets contain more regular traffic samples than attack samples, attempting to replicate real-network traffic. It becomes more challenging for the IDS to recognize specific kinds of attacks as a result, leading to an imbalance in the training data. Our problems thus stem from two factors: unbalanced training data and new, unidentified threats. For an efficient multi-class intrusion detection system, we offer in this paper a hybrid auto encoder-deep neural network (Auto Encoder-DNN) deep learning model that leverages data resampling adaptive synthetic (ADASYN), edited nearest neighbors (ENN) and class weights to get over class imbalance. This advanced technique, Auto Encoder-DNN, focuses on three primary objectives to increase accuracy and performance: 1) lowering the frequencies of false positives and negatives; 2) allowing real-time intrusion detection on fast networks; and 3) detecting zero-day attacks. We evaluate our proposed model Auto Encoder-DNN using the benchmark dataset NSL-KDD, utilizing metrics like as accuracy, precision, recall, and F-score. With an astounding 99.91% accuracy in multi-class classification, the test results show how successful our method is. This proves our IDS's enhanced capacity to defend cloud settings against intrusions by outperforming existing models.