STRATEGY FOR PROTECTING PERSONAL DATA IN MACHINE LEARNING SYSTEMS
DOI:
https://doi.org/10.28925/2663-4023.2025.31.1016Keywords:
mobile inference, neural architecture optimisation, quantisation, pruning, energy-latency trade-off, meta-analysis, ARM processors.Abstract
Massive amounts of personal data drive modern machine- learning pipelines, but that same data can also pose privacy risks. This study gathers and reorganizes scattered empirical evidence on privacy- preserving methods- such as differential privacy, federated optimization, secure aggregation, private transfer learning, and fully homomorphic encryption- into a practical strategy that practitioners can follow confidently. Instead of collecting new datasets, we review twelve peer- reviewed experiments from 2021 to 2025, re- analyze their metrics, and compare the results with regulatory thresholds from GDPR and the draft EU AI Act.
The meta- analysis shows that keeping the privacy budget at two or less maintains macro- F 1 losses under three percentage points across vision, speech, and clinical tasks. However, energy costs increase by a median factor of 2.1. 1. Interestingly, speech- command recognition under DP- SGD became more stable, likely by reducing overfitting. Based on these findings, we introduce a tiered decision matrix: high- sensitivity data require DP- SGD with adaptive clipping; geographically fragmented datasets benefit from federated learning coupled with threshold aggregation; untrusted- cloud deployments need lightweight homomorphic inference; and if none of these apply, private transfer learning on anonymized embeddings remains a solid fallback.
To test the matrix, we use three synthetic but realistic scenarios- critical- care triage, smart- home automation, and retail loyalty prediction- that show how trade- offs change when latency, bandwidth, and legal concerns vary. This framework, called “privacy elasticity,” measures how much model quality can be adjusted before individual rights are at risk and provides practical guidelines for engineers and compliance officers.
By connecting empirical data with ethical principles, this article offers more than just a survey. It presents a coherent theory and an easy- to- use tool. We argue that privacy protection has moved beyond just an add- on feature-
Downloads
References
Taibi, I., & Ramon, J. (2024). Honest fraction differential privacy. Proceedings of the 2024 ACM Workshop on Information Hiding and Multimedia Security (pp. 247–251). ACM. https://doi.org/10.1145/3658664.3659655
Shamsabadi, A. S., Tan, G., Cebere, T. I., Bellet, A., Haddadi, H., Papernot, N., Wang, X., & Weller, A. (2024). Confidential-DPproof: Confidential proof of differentially private training. In 12th International Conference on Learning Representations (ICLR 2024). https://openreview.net/forum?id=PQY2v6VtGe
Xu, R., Li, B., Li, C., Joshi, J. B. D., Ma, S., & Li, J. (2025). TAPFed: Threshold secure aggregation for privacy-preserving federated learning. IEEE Transactions on Dependable and Secure Computing (advance online publication). https://doi.org/10.1109/TDSC.2024.3350206
Byun, S., Sarker, A., Chang, S.-Y., & Byers, B. (2024). Secure aggregation for privacy-preserving federated learning in vehicular networks. ACM Journal on Autonomous Transportation Systems, 1(3), Article 24. https://doi.org/10.1145/3657644
Behnia, R., Chow, S. S. M., Riasi, A., Padmanabhan, B., Ebrahimi, R., & Hoang, T. (2024). e-SeaFL: Efficient secure aggregation for privacy-preserving federated machine learning. In 40th Annual Computer Security Applications Conference (ACSAC ’24) (pp. 135–150). https://arxiv.org/abs/2304.03841
Boscher, C., Benarba, N., Elhattab, F., & Bouchenak, S. (2024). Personalized privacy-preserving federated learning. Proceedings of the 25th ACM/IFIP International Middleware Conference (Middleware ’24) (pp. 348–361). https://doi.org/10.1145/3652892.3700785
Namatevs, I., Sudars, K., Nikulins, A., & Ozols, K. (2025). Privacy auditing in differential private machine learning: The current trends. Applied Sciences, 15(2), 647. https://doi.org/10.3390/app15020647
Liu, Q., Shakya, R., Khalil, M., & Jovanovic, J. (2025). Advancing privacy in learning analytics using differential privacy. In Proceedings of the 15th International Learning Analytics & Knowledge Conference (LAK 2025) (pp. 181–191). ACM. https://doi.org/10.1145/3706468.3706493
Ebel, A., Garimella, K., & Reagen, B. (2025). Orion: A fully homomorphic encryption framework for deep learning. In ASPLOS 2025 – 30th ACM International Conference on Architectural Support for Programming Languages and Operating Systems (pp. 734–749). https://doi.org/10.1145/3676641.3716008
Asghar, H. J., Lu, Z., Zhao, Z., & Kaafar, D. (2025). Practical, private assurance of the value of collaboration via fully homomorphic encryption. Proceedings on Privacy Enhancing Technologies, 2025(2), 258–279. https://doi.org/10.56553/popets-2025-0061
Zhang, F., Zhai, D., Bai, G., Jiang, J., Ye, Q., Ji, X., & Liu, X. (2025). Towards fairness-aware and privacy-preserving enhanced collaborative learning for healthcare. Nature Communications, 16, 2852. https://doi.org/10.1038/s41467-025-58055-3
Haripriya, R., Khare, N., & Pandey, M. (2025). Privacy-preserving federated learning for collaborative medical data mining in multi-institutional settings. Scientific Reports, 15, 12482. https://doi.org/10.1038/s41598-025-97565-4
Published
How to Cite
Issue
Section
License
Copyright (c) 2025 Марія Позднякова
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
