Exploring Hybrid Quantum Classical and Traditional Machining Learning Algorithms for Detecting Cyber Attacks

Abstract

Worldwide internet usage is expanding quickly, creating numerous opportunities in a variety of industries, such as sports, education, entertainment, and finance. Since network technology has grown and become more widely used, it is crucial to manage, maintain, and monitor networks in a bid to maximize economic efficiency and maintain smooth operations. Nonetheless, a major setback of the internet is cyber-attack. This leverages various tools, techniques, and vulnerabilities in systems, social engineering, insider threats, malware and ramsomware to get user credentials and gain access to target network and active assets. This study explores hybrid quantum classical and traditional machining learning algorithms for detecting cyber attack by analyzing the attributes of transmitted packets in benign and scan samples of network traffic data collected by Lawrence Berkeley National Laboratory. Several models such as Hybrid quantum classical model, Convolutional Neural Network, logistic regression, Random Forest, Gradient boosting and support vector classifier were used for training, evaluation and classification of the network traffic data and the results reveal that the hybrid quantum classical modeling terms of time efficiency obtained a training time of 2.36 seconds and evaluation times of 0.02 seconds. In as much as research on quantum computing just began to gain momentum and without a fully functional quantum computer yet in existence, Quantum computing algorithms are already in very close competition with known state-of-the-art algorithms, showing an experimental realization of quantum supremacy over older and existing classical computing.