Quant-Jack: Quantum Machine Learning to Detect Cryptojacking Attacks in IIoT Networks
Pronaya Bhattacharya, Aparna Kumari, Sudeep Tanwar, Ishan Budhiraja, Sahil Patel, Joel J. P. C. Rodrigues · 2024
The paper presents a scheme, Quant-Jack, which employs Quantum Machine Learning (QML) to combat the cryptojacking threat in Industrial Internet-of-Things (IIoT) networks. We propose a dual-layered QML architecture, based on Quantum Neural Networks (QNN) architecture. The first layer is the QNN detection layer that operates via a weighted sum approach on time, frequency, and network traffic. The problem is modeled as multi-objective optimization, which is solved by an iterative Quantum Approximate Optimization Algorithm (QAOA). At the QML filtration layer, a Quantum Metric (QM) is computed that filters anomalies based on specified thresholds. For performance evaluation, the CSE-CIC-IDS 2018 benchmark dataset is augmented with live IIoT data. The performance is simulated for model parameters like convergence rate, attack detection time, network throughput utilization, and other metrics like precision, recall, and F1-score. For 60 nodes, the measured throughput is 11.84 KBps, which is an average improvement of 23.44% compared to baseline quantum models. the QNN model has an accuracy of 0.97 in classifying malign and benign requests, which indicates the efficacy of the proposed scheme against classical security designs.