Search papers, labs, and topics across Lattice.
This study benchmarks quantum machine learning approaches, specifically fidelity-kernel SVMs and variational classifiers, against classical models for detecting cyberattacks in power systems. The research reveals that the evaluation methodology significantly influences the outcomes, with specific choices in the evaluation protocol leading to drastic shifts in performance metrics, such as a macro-F1 score drop from 0.905 to 0.594 based on data splitting methods. The findings underscore the importance of careful evaluation design in machine learning for cybersecurity, as missteps can drastically alter perceived model effectiveness.
Evaluation choices can drastically change the perceived effectiveness of quantum machine learning models for power-system attack detection, revealing a critical vulnerability in benchmarking practices.
Machine-learning detectors for power-system cyberattacks are themselves attack surfaces, and quantum machine learning has been proposed for them. We benchmark fidelity-kernel SVMs and variational classifiers against six tuned classical models on public power-system attack data (Mississippi State/ORNL), across white-box, transfer, decision-based black-box, and poisoning attacks. Our headline finding is methodological: the benchmark's answers are set by the evaluator's choices before the models. Eight choices -- six in the evaluation protocol, two in the tuning the benchmark itself runs -- each reversed or moved a conclusion at fixed models. The largest is the split: the row-level protocol scores 0.905 macro-F1 where holding whole source files out leaves 0.594, and in the capped matched-dimensionality regime the quantum arm sits within noise of chance with the classical arm 0.024 above it. A fidelity kernel looks most robust until attacked directly (retention 0.886 to 0.064); a mis-fitted surrogate manufactures a 10x asymmetry; an unseeded black-box attack moves 75% between restarts. A positive control explains the accuracy null: the labels, not the pipeline. We give the control that catches each choice and release the seeded benchmark.
