Quantum Machine Learning Framework for Early Detection of Cyber Attacks in Critical Infrastructure

Abstract

Due to technological changes leading to increased digitization and networked architecture, critical infras- tructures including electrical power systems, transportation, healthcare and industrial control systems, among others, are more susceptible to highly sophisticated cyber attacks than before. Traditional intrusion detection techniques have been found to be inefficient in identifying attacks because of challenges associated with detecting complex patterns and anomalies in real time. This paper discusses a Quantum Machine Learning (QML) approach to early detection of cyber attacks on critical infrastructure systems. In the approach described herein, quantum feature encoding, quantum circuits, and hybrid quantum and classical optimization processes are exploited for improving detection rate and computational effectiveness. Network traffic and behavior data are represented using quantum states and processed via quantum-inspired machine learning algorithms for detecting attack patterns and cyber intrusions. Experimentation indicates high detection rate, low false positive detections and fast convergence compared to conventional machine learning techniques. Different types of cyber threats are detected including DDoS attacks, malware infections, insider attacks, and Advanced Persistent Threats (APTs).