Differentially Private Deep Learning with Cryptography-Aware Activation Function Design for Edge Intelligence
Keywords:
differential privacy; edge intelligence; cryptography-aware activation functions; secure multiparty computation; homomorphic encryption; federated learning; trustworthy machine learning; socio-technical systemsAbstract
Edge intelligence increasingly depends on deep learning models that operate near data sources, yet this proximity creates difficult tensions among privacy, computation, communication, and trust. Differential privacy offers formal protection for individuals whose data contribute to model training or inference, while cryptographic methods such as secure multiparty computation and homomorphic encryption enable computation over protected inputs. However, these protections interact poorly with conventional activation functions, especially the rectified linear unit, because comparisons, branching, and unbounded piecewise behavior are expensive or impractical in encrypted or distributed settings. This paper examines differentially private deep learning with cryptography-aware activation function design for edge intelligence from a system-level perspective. It argues that activation functions are not merely model-level choices but architectural and governance-relevant components that shape latency, energy consumption, robustness, fairness, auditability, and policy compliance. The discussion synthesizes research on differential privacy, federated learning, secure inference, secure aggregation, and private inference systems. It then develops a layered architecture for edge intelligence that integrates local privacy mechanisms, cryptographic protocols, model compression, and heterogeneous edge infrastructure. The paper analyzes trade-offs among privacy budgets, cryptographic overhead, model accuracy, adversarial resilience, and equitable performance across user groups. It also considers deployment pipelines, regulatory expectations, and sustainability concerns. The conclusion is that cryptography-aware activation design provides a practical but partial route toward trustworthy edge intelligence, requiring coordinated advances in algorithms, hardware, standards, and governance rather than any single technical fix.
References
1. Dwork, C. (2006). Differential privacy. In M. Bugliesi, B. Preneel, V. Sassone, & I. Wegener (Eds.), Automata, Languages and Programming (ICALP 2006), Lecture Notes in Computer Science, Vol. 4052 (pp. 1–12). Springer.
2. Abadi, M., Chu, A., Goodfellow, I., McMahan, H. B., Mironov, I., Talwar, K., & Zhang, L. (2016). Deep learning with differential privacy. In Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security (pp. 308–318). Association for Computing Machinery.
3. Shokri, R., & Shmatikov, V. (2015). Privacy-preserving deep learning. In Proceedings of the 22nd ACM SIGSAC Conference on Computer and Communications Security (pp. 1310–1321). Association for Computing Machinery.
4. McMahan, B., Moore, E., Ramage, D., Hampson, S., & y Arcas, B. A. (2017). Communication-efficient learning of deep networks from decentralized data. In Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (pp. 1273–1282). PMLR.
5. Kairouz, P., McMahan, H. B., Avent, B., Bellet, A., Bennis, M., Bhagoji, A. N., ... & Zhao, S. (2021). Advances and open problems in federated learning. Foundations and Trends in Machine Learning, 14(1–2), 1–210.
6. Juvekar, C., Vaikuntanathan, V., & Chandrakasan, A. (2018). GAZELLE: A low latency framework for secure neural network inference. In Proceedings of the 27th USENIX Security Symposium (pp. 1651–1669). USENIX Association.
7. Liu, J., Juuti, M., Lu, Y., & Asokan, N. (2017). Oblivious neural network predictions via MiniONN transformations. In Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security (pp. 619–631). Association for Computing Machinery.
8. Mohassel, P., & Zhang, Y. (2017). SecureML: A system for scalable privacy-preserving machine learning. In 2017 IEEE Symposium on Security and Privacy (pp. 19–38). IEEE.
9. Riazi, M. S., Weinert, C., Tkachenko, O., Songhori, E. M., Schneider, T., & Koushanfar, F. (2018). Chameleon: A hybrid secure computation framework for machine learning applications. In Proceedings of the 2018 on Asia Conference on Computer and Communications Security (pp. 707–721). Association for Computing Machinery.
10. Mishra, P., Lehmkuhl, R., Srinivasan, A., Zheng, W., & Popa, R. A. (2020). Delphi: A cryptographic inference service for neural networks. In Proceedings of the 29th USENIX Security Symposium (pp. 2505–2522). USENIX Association.
11. Lou, Q., & Jiang, L. (2019). SHE: A fast and accurate deep neural network for encrypted data. In Advances in Neural Information Processing Systems 32 (pp. 10028–10036). Curran Associates.
12. Benaissa, A., Retiat, B., Cebere, B., & Belfedhal, A. E. (2021). TenSEAL: A library for encrypted tensor operations using homomorphic encryption. arXiv preprint arXiv:2104.03152.
13. Papernot, N., McDaniel, P., Sinha, A., & Wellman, M. P. (2018). SoK: Security and privacy in machine learning. In 2018 IEEE European Symposium on Security and Privacy (pp. 399–414). IEEE.
14. Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., & Galstyan, A. (2021). A survey on bias and fairness in machine learning. ACM Computing Surveys, 54(6), Article 115.
15. Peng, Hongwu, et al. "Rrnet: Towards relu-reduced neural network for two-party computation based private inference." arXiv preprint arXiv:2302.02292 (2023).
16. Bonawitz, K., Ivanov, V., Kreuter, B., Marcedone, A., McMahan, H. B., Patel, S., ... & Seth, K. (2017). Practical secure aggregation for privacy-preserving machine learning. In Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security (pp. 1175–1191). Association for Computing Machinery.
17. Bagdasaryan, E., Veit, A., Hua, Y., Estrin, D., & Shmatikov, V. (2020). How to backdoor federated learning. In Proceedings of the 23rd International Conference on Artificial Intelligence and Statistics (pp. 2938–2948). PMLR.
18. Hard, A., Rao, K., Mathews, R., Ramaswamy, S., Beaufays, F., Augenstein, S., ... & Ramage, D. (2018). Federated learning for mobile keyboard prediction. arXiv preprint arXiv:1811.03604.
19. Xu, R., Baracaldo, N., Zhou, Y., Anwar, A., & Ludwig, H. (2019). HybridAlpha: An efficient approach for privacy-preserving federated learning. In Proceedings of the 12th ACM Workshop on Artificial Intelligence and Security (pp. 13–23). Association for Computing Machinery.
20. European Commission. (2021). Proposal for a Regulation laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). COM/2021/206 final.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Computational Intelligence Systems
This work is licensed under a Creative Commons Attribution 4.0 International License.
This article is published under the Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.