Enhancing IoT Security: A Deep Learning Approach with Autoencoder-DNN Intrusion Detection Model
Sarra Cherfi, Ammar Boulaiche, Ali Lemouari · 2024
As the Internet of Things (IoT) continues to proliferate, driving an unprecedented surge in network traffic worldwide, the imperative to safeguard interconnected devices against security threats has become increasingly paramount. In this work, we introduce a new deep learning-based intrusion detection model, specifically by integrating an autoencoder with a deep neural network. We employ a combination of information gain calculation and the simulated annealing algorithm for attribute selection. Evaluation of our model on diverse datasets including UNSW-NB15, TON-IOTwin7, TON-IOTwin10, and CICIDS2017 demonstrates impressive accuracies of 76.90%, 99.90%, 99.86%, and 99.40%, respectively.