An Improved Recurrent Neural Network for Industrial IoT Botnet Attack Detection

Abstract

This research aims to improve the Industrial Internet of Things (IIoT) security, which fosters technological confidence and promotes expansion. The IIoT is mainly used in manufacturing, oil, and gas to avoid botnet attacks. The diverse nature of IIoT devices and limited resources render them susceptible to security breaches. Botnet attacks present serious security risks and compromise IIoT equipment. Several studies applied different learning methods to detect botnet attacks. However, given the unique characteristics of IIoT, obtaining an excellent detection rate with affordable computational needs remains an urgent research problem. A cross-breeding gadget known as long short-term memory (LSTM), an improved gated recurrent unit (GRU), and LSTM-GRU were made to overcome this challenge. An upgraded bi-directional GRU model was developed and applied to the succession of neighborhoods obtained from the LSTM on the info dataset to acquire comprehension of the portrayal. Lastly, the ability to predict the attacks was taught through a supervised learning layer. Investigate the Bot-IoT dataset to approve the adequacy and generalizability of the proposed LSTM-GRU. This model improves the accuracy to 97% and the F1 score to 95% compared to the other existing methods.