Investigation of Benign and Scan Samples of Cyber Security Network Traffic Data: A Hybrid Quantum-Classical Study

Abstract

Worldwide internet usage is expanding quickly, creating numerous opportunities in various industries, such as sports, education, entertainment, and finance. The growth of network technology has necessitated the need for effective management, maintenance, and monitoring of network infrastructures to maximize economic efficiency and maintain smooth operations. Like any coin, the internet has merits and demerits, and cyber attacks are major demerits that every internet user should be aware of. Amongst the most common tactic used in cyber attacks is social engineering, which attackers often leverage to get user credentials and access target networks and active assets. This research employs a hybrid quantum-classical model designed using a simple variational quantum circuit with a single qubit for the training, evaluation, and classification of network traffic data into attacked and non-malicious using a network traffic dataset comprising of benign and scan samples collected by Lawrence Berkeley National Laboratory. The hybrid quantum classical model revealed an accuracy of 99.89% with a training time of 2.36 seconds and evaluation times of 0.02 seconds. Based on these results, it was therefore recommended among others, that the Hybrid Quantum-Classical model holds significant promise for the future of real time anomaly detection in the cyber security space. Overall, the adoption of the Hybrid Quantum-Classical model by cyber security practitioners, could lead to substantial improvements in identifying benign and scan data, thereby enhancing the overall security posture.