Privacy-Preserving Financial Fraud Detection Through Secure Neural Network Inference and Adaptive Polynomial Activation Optimization
Keywords:
privacy-preserving machine learning; secure neural network inference; financial fraud detection; polynomial activation approximation; homomorphic encryption; cost-sensitive calibrationAbstract
Financial fraud detection increasingly depends on neural models that consume detailed transaction records, account histories, and behavioral signals. These data are confidential, and practical deployment therefore requires inference that protects both customer features and proprietary model parameters. Secure neural network inference using homomorphic encryption or multiparty computation offers a principled cryptographic route, but nonlinear activation functions dominate the computational cost and constrain permissible network depth. Fixed low-degree activation replacements can reduce this cost, yet they are usually selected without explicit attention to the cost-sensitive decision boundary of fraud detection or to the multiplicative-depth budget of the secure protocol. This article develops an adaptive polynomial activation optimization framework for privacy-preserving financial fraud detection. The framework treats activation degree, polynomial coefficients, and decision threshold as coupled design variables. It formulates fraud detection as a cost-sensitive risk problem, derives a Chebyshev polynomial family for encrypted evaluation, and optimizes activation approximations under multiplicative-depth and latency constraints. A calibration analysis connects polynomial approximation error to threshold selection and expected fraud cost, while an error-propagation bound characterizes how activation error accumulates across layers. The main insight is that privacy-preserving fraud detection should not minimize activation approximation error alone; it should allocate polynomial capacity according to its effect on fraud cost, score calibration, and secure computation depth. The resulting methodological framework provides a defensible basis for designing secure neural inference systems whose accuracy, privacy, and operational cost are treated as jointly engineered rather than sequentially optimized.
References
1. Abdallah, A., Maarof, M. A., & Zainal, A. (2016). Fraud detection system: A survey. Journal of Network and Computer Applications, 68, 90–113.
2. Dal Pozzolo, A., Caelen, O., Le Borgne, Y.-A., Waterschoot, S., & Bontempi, G. (2014). Learned lessons in credit card fraud detection from a practitioner perspective. Expert Systems with Applications, 41(10), 4915–4928.
3. Phua, C., Lee, V., Smith, K., & Gayler, R. (2010). A comprehensive survey of data mining-based fraud detection research. arXiv preprint arXiv:1009.6119.
4. Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. MIT Press.
5. He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 770–778). IEEE.
6. Acar, A., Aksu, H., Uluagac, A. S., & Conti, M. (2018). A survey on homomorphic encryption schemes: Theory and implementation. ACM Computing Surveys, 51(4), 1–35.
7. Gentry, C. (2009). Fully homomorphic encryption using ideal lattices. In Proceedings of the 41st Annual ACM Symposium on Theory of Computing (pp. 169–178). ACM.
8. Brakerski, Z., Gentry, C., & Vaikuntanathan, V. (2014). (Leveled) fully homomorphic encryption without bootstrapping. ACM Transactions on Computation Theory, 6(3), 1–36.
9. Cheon, J. H., Kim, A., Kim, M., & Song, Y. (2017). Homomorphic encryption for arithmetic of approximate numbers. In Advances in Cryptology – ASIACRYPT 2017 (pp. 409–437). Springer.
10. Gilad-Bachrach, R., Dowlin, N., Laine, K., Lauter, K., Naehrig, M., & Wernsing, J. (2016). CryptoNets: Applying neural networks to encrypted data with high throughput and accuracy. In Proceedings of the 33rd International Conference on Machine Learning (pp. 201–210). PMLR.
11. 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.
12. 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.
13. 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. 4494–4503). PMLR.
14. Peng, H., Huang, S., Zhou, T., Luo, Y., Wang, C., Wang, Z., ... & Ding, C. (2023, October). Autorep: Automatic relu replacement for fast private network inference. In 2023 IEEE/CVF International Conference on Computer Vision (ICCV) (pp. 5155-5165). IEEE.
15. Hesamifard, E., Takabi, H., & Ghasemi, M. (2017). CryptoDL: Deep neural networks over encrypted data. arXiv preprint arXiv:1711.05189.
16. Nair, V., & Hinton, G. E. (2010). Rectified linear units improve restricted Boltzmann machines. In Proceedings of the 27th International Conference on Machine Learning (pp. 807–814). Omnipress.
17. 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.
18. Bost, R., Popa, R. A., Tu, S., & Goldwasser, S. (2015). Machine learning classification over encrypted data. In Proceedings of the 22nd Network and Distributed System Security Symposium. Internet Society.
19. Chou, E., Beal, J., Levy, D., Yeung, S., Haque, A., & Fei-Fei, L. (2018). Faster CryptoNets: Leveraging sparsity for real-world encrypted inference. arXiv preprint arXiv:1811.09953.
20. Lou, Q., & Jiang, L. (2019). SHE: A fast and accurate deep neural network for encrypted data. In Advances in Neural Information Processing Systems 32. Curran Associates.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Journal of Advanced Financial Research

This work is licensed under a Creative Commons Attribution 4.0 International License.