A Resilient Cybersecurity Framework for IIoT Systems Using AE and RNN-Based Threat Detection

Abstract

Security in the Industrial Internet of Things system is of utmost importance because such systems are being integrated more and more into very important industries like manufacturing, health, transportation, and energy. With the proliferation of more connected devices and growing sensor data exponentially, IIoT systems become the prime target for cyberattacks that threaten data integrity and operational continuance of industries. This paper proposes an efficient AE+RNN-based security model for the detection of cyber-attacks in IIoT systems. The proposed model leverages the strengths of Autoencoders in feature extraction with the power of Recurrent Neural Networks in learning temporal sequences and hence detects anomalies and complex attack patterns effectively. The AE reduces high dimensionality of voluminous sensor data, which is further fed into an RNN that can proficiently capture sequential dependencies associated with time-sensitive environments and identify threats. The proposed model's performance has been compared to five state-of-the-art techniques, namely CNN, LSTM, Random Forest, SVM, and Hybrid DL. It also has the possibility of detecting wide-ranging cyber-attacks, botnet-driven DDoS, and hence could act as an efficient and effective tool towards improving the security of IIoT. Conclusion: This paper has represented the AE+RNN-based model as a promising solution for the growing cybersecurity challenges of IIoT systems and has provided an efficient, scalable, and adaptive approach toward the detection and mitigation of threats in real time.