Performance analysis of recurrent neural networks for intrusion detection systems in Industrial-Internet of Things
Mohammed Tayebi, Said El Kafhali
Abstract
As the world witnesses significant technological advancements, it faces complex challenges threatening these innovations. Industry 4.0 has emerged as a pivotal development of the century, offering advanced technological solutions to enhance human life and create a highly connected world. So, securing those systems is a necessity. In this paper, we present a performance analysis of various recurrent neural network (RNN) architectures, namely recurrent neural networks (RNNs), long-short-term memory networks (LSTMs), gated recurrent units (GRUs), bidirectional long-short-term memory networks (BiLSTMs), and bidirectional gated recurrent units (BiGRUs) to identify the most effective model for intrusion detection systems (IDS) in IoT environments. To enhance the performance of these architectures, we propose the integration of a deep neural network (DNN) and evaluate the impact of adding attention mechanisms. The models were tested on two benchmark datasets to comprehensively assess their performance in terms of accuracy, precision, recall, F1 score, and training time. The results show that the DNN-enhanced architectures consistently outperformed their baseline and attention models, achieving the best overall performance across all datasets.