Quantum Machine Learning for Cyber-Physical Anomaly Detection in Unmanned Aerial Vehicles: A Leakage-Free Evaluation with Proxy-Audited Feature Sets
Researchers have developed a novel approach using quantum machine learning to detect anomalies in unmanned aerial vehicles (UAVs). The study introduces a leakage-free evaluation method on the TLM:UAV benchmark, employing a group-aware temporal protocol and a three-mode feature audit to assess the impact of different signal types. While a hybrid XGBoost and Data Reuploading (DRU) classifier showed incremental benefits, its statistical significance was limited by inter-seed variance, though it achieved the lowest false-alarm rate under proxy-free conditions. The team has released an open-source Qiskit implementation for cybersecurity analytics in NISQ-era aerospace systems. AI
IMPACT This research offers a potential pathway for improving cybersecurity in UAVs by exploring quantum machine learning for anomaly detection.