Hardware-Aware Neural Network Optimization for Secure Two-Party Inference Using Lightweight Activation Functions
Keywords:
secure two-party computation; private inference; hardware-aware optimization; lightweight activation functions; privacy-preserving machine learning; neural architecture search; deployment governance; socio-technical systemsAbstract
Secure two-party inference has emerged as a critical capability for organizations that need to evaluate neural networks over sensitive inputs without revealing either the model or the data to the other participant. Although cryptographic protocols provide strong formal guarantees, their practical deployment remains constrained by communication overhead, latency, hardware heterogeneity, and the high cost of nonlinear activation functions. This paper examines hardware-aware neural network optimization for secure two-party inference, with particular attention to lightweight activation functions as a structural design lever. The analysis treats private inference not simply as a cryptographic problem but as a socio-technical infrastructure problem involving model architecture, accelerator selection, network topology, deployment governance, and long-term operational sustainability. We synthesize insights from secure computation, efficient deep learning, hardware-aware architecture search, and privacy-preserving machine learning systems. The discussion shows that activation function choice interacts with protocol design, memory hierarchy, batching strategy, and threat models in ways that cannot be resolved through isolated algorithmic improvements. Lightweight activation functions can reduce the number of expensive nonlinear operations, but they may also alter model accuracy, fairness, robustness, and auditability. The paper argues that sustainable private inference requires co-design across layers: model designers should treat activation functions as both accuracy parameters and cryptographic cost parameters; system architects should align protocols with hardware capabilities; and policymakers should require transparency about residual leakage, failure modes, and equitable access. The conclusion outlines a research agenda for hardware-aware, activation-conscious private inference that balances performance, privacy, fairness, and institutional accountability.
References
1. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). ImageNet classification with deep convolutional neural networks. Advances in Neural Information Processing Systems, 25, 1097-1105.
2. LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444. https://doi.org/10.1038/nature14539
3. Sze, V., Chen, Y.-H., Yang, T.-J., & Emer, J. S. (2017). Efficient processing of deep neural networks: A tutorial and survey. Proceedings of the IEEE, 105(12), 2295-2329. https://doi.org/10.1109/JPROC.2017.2761740
4. Yao, A. C. (1986). How to generate and exchange secrets. In Proceedings of the 27th Annual Symposium on Foundations of Computer Science (pp. 162-167). IEEE. https://doi.org/10.1109/SFCS.1986.25
5. Goldreich, O., Micali, S., & Wigderson, A. (1987). How to play any mental game. In Proceedings of the 19th Annual ACM Symposium on Theory of Computing (pp. 218-229). ACM. https://doi.org/10.1145/28395.28420
6. 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. https://doi.org/10.1109/SP.2017.12
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). ACM. https://doi.org/10.1145/3133956.3134056
8. Rouhani, B. D., Riazi, M. S., & Koushanfar, F. (2018). DeepSecure: Scalable provably-secure deep learning. In Proceedings of the 55th Annual Design Automation Conference (pp. 1-6). ACM. https://doi.org/10.1145/3195970.3196023
9. Peng, H., Huang, S., Zhou, T., Luo, Y., Wang, C., Wang, Z., Zhao, J., Xie, X., Li, A., Geng, T., Mahmood, K., Ding, C., & Wen, W. (2023). Rrnet: Towards relu-reduced neural network for two-party computation based private inference. arXiv. https://arxiv.org/abs/2302.02292
10. Rathee, D., Rathee, M., Kumar, N., Chandran, N., Gupta, D., Rastogi, A., & Sharma, R. (2020). CrypTFlow2: Practical 2-party secure inference. In Proceedings of the 2020 ACM SIGSAC Conference on Computer and Communications Security (pp. 325-342). ACM. https://doi.org/10.1145/3372297.3417274
11. 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.
12. Patra, A., Schneider, T., Suresh, A., & Yalame, H. (2020). BLAZE: Blazing fast privacy-preserving machine learning. In Proceedings of the 2020 Network and Distributed System Security Symposium. Internet Society. https://doi.org/10.14722/ndss.2020.24098
13. Knott, B., Venkataraman, S., Hannun, A., Sengupta, S., & van der Maaten, L. (2021). CrypTen: Secure multi-party computation meets machine learning. In Advances in Neural Information Processing Systems 34 (pp. 4961-4973).
14. Dathathri, R., Saarikivi, O., Chen, H., Laine, K., Lauter, K., Maleki, S., & Musuvathi, M. (2019). CHET: An optimizing compiler for fully-homomorphic neural-network inferencing. In Proceedings of the 40th ACM SIGPLAN Conference on Programming Language Design and Implementation (pp. 142-156). ACM. https://doi.org/10.1145/3314221.3314628
15. Chillotti, I., Gama, N., Georgieva, M., & Izabachène, M. (2020). TFHE: Fast fully homomorphic encryption over the torus. Journal of Cryptology, 33(1), 34-91. https://doi.org/10.1007/s00145-019-09319-x
16. Cai, H., Gan, C., Wang, T., Zhang, Z., & Han, S. (2020). Once-for-all: Train one network and specialize it for efficient deployment. In International Conference on Learning Representations.
17. Han, S., Mao, H., & Dally, W. J. (2016). Deep compression: Compressing deep neural networks with pruning, trained quantization and Huffman coding. In International Conference on Learning Representations.
18. Sanyal, A., Kusner, M. J., Gascón, A., & Kanade, V. (2018). TAPAS: Tricks to accelerate (encrypted) prediction as a service. In Proceedings of the 35th International Conference on Machine Learning (pp. 4490-4499). PMLR.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 International Journal of Artificial Intelligence Engineering and 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.
