J. Ludmir and T. Patel
Rice University, Texas, United States
Keywords: Quantum Computing, Anomaly Detection, Energy, Defense, Machine Learning
This work introduces a quantum machine learning-based approach for anomaly detection with significant potential for applications in critical sectors such as energy and defense. In these sectors, detecting anomalies is crucial for identifying system failures, cyber threats, or unauthorized access, which can have far-reaching consequences. Traditional anomaly detection methods like clustering and Isolation Forests struggle with complex, high-dimensional data often present in energy grid monitoring and cybersecurity systems. The work leverages quantum autoencoders to enhance the detection of subtle, hard-to-detect anomalies by compressing and reconstructing quantum representations of data. Quantum embedding techniques allow for the efficient handling of large datasets by encoding classical information into quantum states, enabling superior pattern recognition capabilities. The optimization process uses a dual annealing algorithm to fine-tune gate parameters, improving the system’s ability to detect anomalies. By progressively compressing data and refining gate operations, the model achieves significant performance gains. The work demonstrates over 94% accuracy in detecting anomalies and highlights the system's ability to distinguish anomalies based on reconstruction errors. This quantum-based approach offers transformative possibilities for real-time, scalable anomaly detection in energy and defense, where rapid and accurate identification of outliers can prevent system failures, enhance security, and improve resilience.