M-TADS: A Multi-Trust DoS Attack Detection System for MEC-enabled Industrial loT

Abstract

The Industrial Internet of Things (IIoT) remains an inevitable system in various applications that require data collection and processing in the modern industrial revolution. The IIoTs are responsible for critical data collection and transmission to cloud servers to address life-dependent problems. However, these cyber-physical devices are vulnerable to network attacks such as selective forwarding, flooding, and Sybil attacks. Meanwhile, behavioural patterns characterise the IIoT devices under such attacks due to their effect on transmission latency, power consumption, and computational time. Hence, this paper presents a multi-trust security system to monitor and record these parameters, such as network byte-in and byte-out, CPU usage, and energy consumption on the IIoT device. Based on the ML model, we created an efficient multi-trust attack detection system (M-TADS) to detect denial of service attacks (DoS) in the IIoT. IIoT devices have resource constraints that practically prevent them from fully implementing the proposed M-TADS on the same cyber-physical device. Hence, the captured parameters from the IIoT devices are offloaded to a deep neural network model created with long short term memory (LSTM). The LSTM is hosted on a multi-access edge computing (MEC) server at the network edge to determine the possible existence of the DoS attack signature. Due to the high latency accompanying DoS attack, we introduce a custom hold and check filter on the IIoT devices. The proposed M-TADS performance is verified through simulations, and the results confirm high performance in terms of throughput, energy consumption, packet delay, and IIoT network DoS attack detection accuracy.