RF Spectrogram Anomaly Detection with Quantum Kitchen Sinks: Architecture, Representation, and Hardware Validation

TL;DR — This paper introduces an extended Quantum Kitchen Sinks (QKS) hybrid quantum-classical feature map for RF spectrogram anomaly detection, validated on real quantum hardware. It matters as a reproducible, empirically grounded framework for applying near-term quantum ML methods to a practical security task.

• Extends standard QKS with multi-depth data re-uploading and ring entanglement, evaluated via a "validation-locked" five-stage ablation isolating architecture, re-uploading depth, episode budget, input representation, and classical readout.
• DCT input representations consistently beat raw and PCA inputs; moderate-depth entangled QKS configs form the strongest regime, and QKS beats matched classical direct-readout baselines across all representation-readout pairs.
• Best configuration reaches test AUROC 0.8778 and F1 0.7995 on held-out data.
• Bridges real-world realism on both ends: real measured sub-6 GHz cellular signals plus real-device runs on the ibm_quebec QPU, with AUROC deviations below 0.013 versus simulation.