Intrusion Detection in IoT Networks Using Deep Learning: A Comprehensive Approach
Abdallah S. Elnamaky, Emad A. Elsamahy, Samir G. Sayed, Ahmed Salem · 2025
Due to the wide availability and usage of Internet of Things (IoT) devices in many fields, these devices have inherent vulnerabilities due to their design. Therefore, the probability of cyberattacks which target the IoT networks has significantly increased. To increase the security of IoT networks, this study suggests an Intrusion Detection System (IDS) that makes use of a deep learning technique. Moreover, to effectively categorize network traffic into normal and potentially harmful attacks, our suggested approach utilizes deep learning techniques such as: Long Short-Term Memory (LSTM), Recurrent neural network (RNN), Dense neural network (DNN). Reducing the dimensionality of traffic characteristics using feature selection approaches minimizes the detection delays and computational resources. The system was trained and tested using benchmark BoT-IoT dataset which has a variety of attack types. Our findings demonstrate that the suggested IDS attain high accuracy and F1-score of 96.91% and 95.78% in addition to low false negative and false positive rates 0.1690 and 0.00834 in comparison to their corresponding results in literature.