An enhanced security mechanism for internet of things using deep learning

Abstract

The Internet of Things (IoT) has revolutionized connectivity, but it also introduces significant security challenges. This thesis presents an enhanced security mechanism for IoT using advanced deep learning techniques to develop a robust Intrusion Detection System (IDS). The research is divided into two primary models: the Hybrid Optimized Learning Model for Intrusion Classification and the Hybrid Model for Intrusion Detection. The Hybrid Optimized Learning Model combines Salp Swarm Optimization (SSO) and Bee Foraging Optimization (BFO) for efficient feature selection and employs Multiplicative Long Short-Term Memory (MLSTM) for accurate intrusion classification. The second model integrates AlexNet for feature extraction, Bidirectional LSTM (BiLSTM) for optimal feature selection, and Decision Tree (C5.0) for final classification. Experimental results demonstrate superior performance of the proposed models in terms of accuracy, precision, recall, and F1-score compared to existing IDS solutions. The integration of nature-inspired optimization algorithms with deep learning enhances the detection capabilities, addressing IoT-specific challenges such as resource constraints and high data variability. This research contributes to the development of scalable and efficient security solutions, ensuring the protection of IoT networks against diverse cyber threats. newline