A Comprehensive Analysis of Intrusion Detection in Internet of Things (IoT)

The intrusion detection systems (IDS) play a vital role in both identifying malicious activity and enhancing network security. An IDS must be introduced to mitigate and identify hostile attacks within networks. In recent years, IDS has become a critical issue in the cyber security of Machine Learning, and Deep Learning techniques have been applied to IDS to increase their efficiency and accuracy. In this paper, several methods such as Stacked Contractive Autoencoder with Support Vector Machine, Convolutional Neural Network with Bidirectional Long Short-Term Memory, Particle Swarm Optimization with Convolutional Neural Network, and Convolutional Neural Network are used in the feature extraction process. Second, the Trust-based Intrusion Detection and Classification System, Modified version of Growth Optimizer, Aquila Optimizer, and Transient Search Optimizer methods are used in this process for feature selection. Finally, K-means with Random Forest, Single Hidden Layer Feed-Forward Neural Network, and Decision Tress methods are utilized in the classification. Extensive evaluation and comparisons of the proposed approaches were conducted by utilizing public datasets from cloud and IoT settings. The methods employed yielded outstanding outcomes in terms of identifying previously undiscovered attacks with a high degree of precision.