Levy Flight Strategy-based Tasmanian Devil’s Optimization and Autoencoder Approach for Intrusion Detection in Cloud Computing

Cloud Computing (CC) is an Internet-based technology that offers shared resources such as storage, software and platforms to users on demand.However, Intrusion Detection System (IDS) often encounter challenges when processing high-dimensional data, resulting in low accuracy and increased false alarm rate, which contribute to unreliable model predictions.To enhance the accuracy and reliability of IDS in processing high-dimensional information, it is essential to address errors in the training data.This research proposes a Levy Flight strategy-based Tasmanian Devil's Optimization (LFS-TDO) approach for feature selection and an Autoencoder (AE) for classifying network attacks.The LFS-TDO strategy demonstrates a significant improvement in optimization performance, where the small step size in levy flight enhances the exploitation ability.The AE mechanism effectively captures complex and nonlinear relationships in data, making it highly effective for detecting subtle anomalies and variations in network traffic.The proposed LFS-TDO-based AE technique is evaluated using UNSW-NB15, NSL-KDD and CICIDS2017 datasets, achieving accuracies of 93.81%, 99.82%, 99.70%, respectively, with precision scores of 99.75%, 99.80%, 99.78%, recall rates of 92.28%, 99.76%, 97.35%, and F1-scores of 98.95%, 99.78%, 99.80, respectively.These results demonstrate a significant improvement over existing techniques like Convolutional Neural Network (CNN) and Genetic Algorithm and Particle Swarm Optimization (GA-PSO) approach.