Trustworthy and Reliable Deep Learning-based Cyberattack Detection in Industrial IoT.

Area of Science:

• Computer Science
• Cybersecurity
• Network Engineering

Background:

• Industrial Internet of Things (IIoT) networks require high trustworthiness and sustainability for critical tasks.
• Traditional security mechanisms are inadequate for IIoT due to protocol differences and outdated adaptations.
• Novel approaches are essential to enhance security, privacy, and trust in IIoT networks.

Purpose of the Study:

• To propose a novel approach for improving the trustworthiness of IIoT-enabled networks.
• To develop an accurate and reliable cyberattack detection scheme for Supervisory Control and Data Acquisition (SCADA) networks within IIoT.
• To enhance security and privacy mechanisms in IIoT environments.

Main Methods:

• A hybrid deep learning model combining Pyramidal Recurrent Units (PRU) and Decision Tree (DT) was developed.
• An ensemble-learning method was employed for cyberattack detection in SCADA-based IIoT networks.
• The PRU's non-linear learning and ensemble DT addressed feature sensitivity for high detection rates.

Main Results:

• The proposed scheme demonstrated superior performance compared to traditional and existing machine learning detection methods.
• High detection rates were achieved by effectively handling irrelevant features.
• The approach significantly improved the security and trustworthiness of IIoT-enabled networks.

Conclusions:

• The novel PRU and ensemble DT-based cyberattack detection scheme offers a robust solution for IIoT security.
• The proposed method enhances the reliability and trustworthiness of SCADA networks in industrial settings.
• This research contributes to securing critical IIoT infrastructure against cyber threats.