Equivariant Quantum Clustering with Differential Privacy: Parameter-Efficient Privacy-Preserving Analysis Across Heterogeneous Sensitive Datasets
Journal of Ai ML DL · 01.01.2025 · 1 (1) · 1-24
DOI: 10.25163/ai.1110790 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Ai ML DL, online erschienen am 01.01.2025.
Abstract (Originalsprache)
Clustering sensitive data — patient records, network traffic logs, behavioral profiles — sits at an uncomfortable intersection of analytical necessity and privacy risk. Traditional clustering algorithms were not designed with privacy in mind, and adapting them retrospectively through differential privacy noise typically degrades clustering utility in ways that limit practical value. Quantum computing offers a structurally different computational substrate, but whether its theoretical privacy advantages translate into measurable empirical gains remains an open and genuinely contested question. This paper introduces Symmetry-Aware Equivariant Quantum Clustering (EQC), a framework that integrates p4m symmetry constraints into quantum circuits via parameter sharing, combined with rigorously composed differential privacy guarantees across all pipeline stages. Three privacy-sensitive datasets were used for evaluation — NSL-KDD network intrusion records, CERT Insider Threat v6.2 behavioral logs, and a Synthetic MIMIC-III clinical dataset — spanning meaningfully different domain characteristics. EQC achieves 79.3% clustering accuracy on NSL-KDD while reducing membership inference attack success to 38.3%, compared to 61.7% for classical baselines under equivalent privacy budgets (ε = 1.0, δ = 10?5). Ablation studies confirm that performance gains arise primarily from parameter reduction through equivariance constraints and differential privacy noise rather than from uniquely quantum mechanical effects — an honest finding that reshapes, but does not diminish, the contribution. EQC establishes a credible quantum-ready clustering framework with rigorously validated privacy-utility tradeoffs, designed for sensitive data environments where both formal guarantees and practical accuracy are non-negotiable.
Zitieren
APA
(2025). Equivariant Quantum Clustering with Differential Privacy: Parameter-Efficient Privacy-Preserving Analysis Across Heterogeneous Sensitive Datasets. Journal of Ai ML DL, 1(1), 1-24. https://doi.org/10.25163/ai.1110790
Vancouver
Equivariant Quantum Clustering with Differential Privacy: Parameter-Efficient Privacy-Preserving Analysis Across Heterogeneous Sensitive Datasets. Journal of Ai ML DL. 2025;1(1):1-24. doi:10.25163/ai.1110790
BibTeX
@article{euvimed2025,
title = {Equivariant Quantum Clustering with Differential Privacy: Parameter-Efficient Privacy-Preserving Analysis Across Heterogeneous Sensitive Datasets},
journal = {Journal of Ai ML DL},
year = {2025},
volume = {1},
number = {1},
pages = {1--24},
doi = {10.25163/ai.1110790}
}
EUVIMED