Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM68351
Volume:
10 |
10 |
Oct 2026
Pages:
1-9
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
Computer Engineering
Sandip Institute of Technology and Research Centre (SITRC)
Computer Engineering
Sandip Institute of Technology and Research Centre (SITRC)
Computer Engineering
Sandip Institute of Technology and Research Centre (SITRC)
Computer Engineering
Sandip Institute of Technology and Research Centre (SITRC)
Computer Engineering
Sandip Institute of Technology and Research Centre (SITRC)
The rapid adoption of machine learning across sensitive domains—such as digital healthcare, automated financial auditing, and biometric security—has highlighted the chal-lenge of balancing high predictive utility with strict data privacy. Traditional Privacy-Preserving Machine Learning (PPML) techniques rely primarily on cryptographic mecha-nisms, such as Fully Homomorphic Encryption, or perturba-tion techniques, such as Differential Privacy. However, cryptographic frameworks incur high computational and latency overheads, while noise-injection methods degrade model utility and classification performance. To resolve this privacy-utility trade-off, deep representation learning has emerged as an architectural alternative. This literature survey provides an exhaustive review of robust representa-tion learning techniques designed for privacy-preserving machine learning. It traces the transition from basic unsu-pervised autoencoders to supervised multi-objective resid-ual autoencoders, analyzing how intermediate feature ag-gregation, center loss constraints, and subspace alignment functions preserve task-relevant discriminative signals while stripping explicit sensitive descriptors. Furthermore, this survey explores the operationalization of representa-tion learning across both unimodal cloud outsourcing and vertically distributed multimodal systems, evaluates system vulnerability profiles against advanced Model Inversion and Membership Inference Attacks, and synthesizes key empirical benchmarks across high-dimensional biological, clinical, and vision datasets. Finally, critical open challeng-es—including supervision dependence, horizontal federat-ed adaptation, and system explainability —are identified to chart future research directions.
KEYWORDS
This survey has comprehensively examined robust representation learning as an architectural approach to privacy-preserving machine learning. By transforming raw inputs into multi-layer concatenated latent embeddings through multi-objective residual autoencoders, representation learning resolves the privacy-utility trade-off inherent in conventional cryptographic and noise-injection paradigms. Integrating intermediate feature aggregation, Center Loss constraints, and linear PCA subspace alignment ensures that shared embeddings preserve discriminative task-relevant features while obfuscating explicit raw descriptors. Empirical evaluations across clinical imaging, high-dimensional genomic microarrays, and multimodal multi-omics datasets demonstrate that multi-objective representation learning matches or exceeds raw model classification accuracy while providing strong empirical defense against Model Inversion and Membership Inference Attacks.
1. R. Shokri, M. Stronati, C. Song, and V. Shmatikov, "Membership inference attacks against machine learning models," in Proc. IEEE Symp. Security and Privacy (S&P), 2017, pp. 3–18. 2. M. Fredrikson, S. Jha, and T. Ristenpart, "Model inversion attacks that exploit confidence information and basic countermeasures," in Proc. ACM SIGSAC Conf. Computer and Communications Security (CCS), 2015, pp. 1322 –1333. 3. K.-C. Wang, Y. Fu, K. Li, A. Khisti, R. Zemel, and A. Makhzani, "Variational model inversion attacks," in Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 34, 2021. 4. D. Ye, T. Zhu, S. Shen, and W. Zhou, "Label-only model inversion attacks: Attack with the least information," IEEE Trans. Inf. Forensics Security, 2022. 5. N. Rochel, "Ethical and legal perspectives on GDPR compliance," Digital Policy Studies, 2021. 6. J. Lee, E. Lee, J.-W. Lee, Y. Kim, Y.-S. Kim, and J.-S. No, "Privacy-preserving machine learning with fully homomorphic encryption for deep neural networks," IEEE Access, 2022. 7. J. Lee et al., "Precise approximation of convolutional neural networks for homomorphically encrypted data," IEEE Access, 2023. 8. L. F. Quintero-Ossa et al., "Enhancing privacy-preserving data sharing via autoencoder latent embeddings," in Proc. Int. Conf. Data Science, 2022. 9. A. Ferreira, D. Silva, and R. Costa, "Autoencoder-based collaborative filtering for recommendation systems," in Proc. ACM Recommender Systems Conf., 2020. 10. Y. Pan, F. He, and H. Yu, "Social recommendation with deep autoencoder-based representation learning," Neurocomputing, 2020.
Publication Date
10/26/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
1
Total views since publication
PDF Downloads
5
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...