A Comparative Study of IDS-Based Deep Learning Models for IoT Network
Bassam Noori Shaker, Bahaa Al-Musawi, Mohammed Falih Hassan
Abstract
The proliferation of connected devices within Internet of Things (IoT) networks has underscored the critical necessity of developing efficient Intrusion Detection Systems (IDS) that can adeptly identify and counteract threats. Deep learning algorithms are better than classical machine learning at automatically learning patterns and representations from raw data, enabling them to detect complex attacks. This research aims to conduct a comparative study of IDSs implemented through three deep learning models: Convolutional Neural Networks (CNN), Deep Neural Networks (DNN), and Recurrent Neural Networks (RNN). The study employs three datasets (NF-UNSW-NB15, NF-BoT-IoT, and NF-ToN-IoT) to assess model performance in both binary and multiclass classification scenarios. The findings indicate that for binary traffic classification, DNN outperforms the other two models with an accuracy bout (98%) for all datasets. In the context of multiclass traffic classification, the DNN model surpasses the performance of the other model except for NF-ToN-IoT a CNN model score (69.08%).