Journals / Afyon Kocatepe Üniversitesi Fen ve Mühendislik Bilimleri Dergisi / 2026 / Cilt: 26 Sayı: 1

Homomorphic Encryption in Finance: Training and Inference on Encrypted Data with Concrete ML

Finansta Homomorfik Şifreleme: Concrete ML ile Şifrelenmiş Veriler Üzerinde Eğitim ve Çıkarım

Pages
64–77
Publication Date
2026-01-19
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

The findings obtained indicate that this approach developed with Concrete ML provides a secure and effective solution in financial applications where privacy is critical.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:Although machine learning is a significant technology that offers useful insights into financial analysis procedures, the actual use of these technologies is restricted by data privacy concerns.
📊
Key Findings & Data:A completely homomorphic encryption-based approach presented in this paper to train machine learning models and generate predictions while protecting the privacy of business financial data.
💡
Conclusion & Impact:The findings obtained indicate that this approach developed with Concrete ML provides a secure and effective solution in financial applications where privacy is critical.
🔬Design:Modeling & Simulation
📍Scope:Comprehensive Sample
⚙️Variables:Homomorfik Şifreleme · Finansal Risk Tahmini · Concrete ML Çerçevesi

Abstract

Although machine learning is a significant technology that offers useful insights into financial analysis procedures, the actual use of these technologies is restricted by data privacy concerns. A completely homomorphic encryption-based approach presented in this paper to train machine learning models and generate predictions while protecting the privacy of business financial data. The goal of the suggested approach is to use encrypted data to generate predictions while training machine learning models on open data.During the training stage, financial ratios from publicly accessible sources were used to train machine learning models on open data and prepare them for encrypted inference.The client-side encrypted financial information is transmitted to the server during the inference phase, and the prediction procedure was conducted in an encrypted. During the inference phase, the client-side financial data was encrypted and transmitted to the server. The prediction process was also encrypted, and the outcome was only decrypted on the client side. The Altman Z-Score, L-Score, and Zmijewski Score models were employed in the research, and the encrypted models' accuracy values (R2: 0.68-0.98) performed comparably to the open models. The accuracy levels of the encrypted models have produced results that are quite similar to those of the open models in experiments utilizing models predicting a company's status, such as the Altman Z-Score, L-Score, and Zmijewski Score. Performance metrics including inference and training times have also been examined. The findings obtained indicate that this approach developed with Concrete ML provides a secure and effective solution in financial applications where privacy is critical.

Özet

Makine öğrenimi, finansal analiz süreçlerine faydalı bilgiler sunan önemli bir teknoloji olmasına rağmen, bu teknolojilerin gerçek kullanımı veri gizliliği endişeleriyle sınırlıdır. Bu makalede, iş dünyasının finansal verilerinin gizliliğini koruyarak makine öğrenimi modellerini eğitmek ve tahminler oluşturmak için tamamen homomorfik şifreleme tabanlı bir yaklaşım sunulmuştur. Önerilen yaklaşımın amacı, açık veriler üzerinde makine öğrenimi modellerini eğitmek ve şifrelenmiş verilerle tahminler oluşturulurken veri gizliliğini sağlamaktır. Eğitim aşamasında, kamuya açık kaynaklardan alınan finansal oranlar kullanılarak makine öğrenimi modelleri eğitilmiş ve şifrelenmiş çıkarım için hazır hale getirilmiştir. Çıkarım aşamasında, istemci tarafındaki şifrelenmiş finansal bilgiler sunucuya iletilmiş ve tahmin işlemi şifrelenmiş olarak yapılmıştır. Çıkarım aşamasında, istemci tarafındaki finansal veriler şifrelenmiş ve sunucuya iletilmiştir. Tahmin süreci de şifrelenmiş olup, sonuç yalnızca istemci tarafında deşifre edilmiştir. Araştırmada Altman Z-Score, L-Score ve Zmijewski Score modelleri kullanılmış ve şifrelenmiş modellerin doğruluk değerleri (R2: 0.68-0.98), açık modellerle karşılaştırılabilir şekilde performans göstermiştir. Şifrelenmiş modellerin doğruluk düzeyleri, Altman Z-Score, L-Score ve Zmijewski Score gibi bir şirketin durumunu tahmin eden modellerle yapılan deneylerde, açık modellerle benzer sonuçlar üretmiştir. Ayrıca, çıkarım ve eğitim süreleri gibi performans metrikleri de incelenmiştir. Elde edilen bulgular, Concrete ML ile geliştirilen bu yaklaşımın, finansal uygulamalarda gizliliğin kritik olduğu durumlarda güvenli ve etkili bir çözüm sunduğunu göstermektedir.

🔖

Cite This Article

Yilmaz, E., Aktürk, B., Civelek, D., Büyüktanır, T., Yıldız, K. (2026). Homomorphic Encryption in Finance: Training and Inference on Encrypted Data with Concrete ML. Afyon Kocatepe Üniversitesi Fen ve Mühendislik Bilimleri Dergisi, Cilt: 26 Sayı: 1, 64–77. https://doi.org/10.35414/akufemubid.1686923

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (29)

  1. Ahamed, S.I. and Vadlamani, R., 2022. Privacy-Preserving Wavelet Neural Network with Fully Homomorphic Encryption. arXiv preprint, arXiv:2205.13265. https://doi.org/10.48550/arXiv.2205.13265
  2. Bradley, S., 2024. Comparative Analysis of Two Data Privacy Regulatory Schemes: The GDPR and the CCPA. Studentworks, eRepository, Seton Hall University, USA Brickell, E.F. and Yacobi, Y., 1988. On privacy homomorphisms (extended abstract). In Proceedings of the Workshop on Theory and Applications of Cryptographic Techniques, 117–125.
  3. Basu, P. Roy, T. S., Naidu, R., & Muftuoglu, Z. l., 2021. Privacy enabled financial text classification using differential privacy and federated learning. https://doi.org/10.48550/arXiv.2110.01643
  4. Chen, T. and Guestrin, C., 2016. XGBoost: A scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794.
  5. Evren, A.G., 2023. Comparative Analysis of the European Union and Turkish Personal Data Protection Laws: Basic Principles, Legal Grounds, and Rights of Data Subjects. Kişisel Verileri Koruma Dergisi, 5(2), 39–64.
  6. Faneela, B.G., Ahmad, J., Buchanan, W.J. and Jan, S.U., 2025. Privacy-Preserving Credit Card Approval Using Homomorphic SVM: Toward Secure Inference in FinTech Applications. arXiv preprint arXiv:2505.05920. https://doi.org/10.48550/arXiv.2505.05920
  7. Frery, J., Stoian, A., Bredehoft, R., Montero, L., Kherfallah, C., Chevallier-Mames, B., & Meyre, A., 2023. Privacy-preserving tree-based inference with fully homomorphic encryption. Cryptology ePrint Archive. https://doi.org/10.48550/arxiv.2303.01254
  8. Friedman, J.H., 2001. Greedy function approximation: A gradient boosting machine. Annals of Statistics, 29(5), 1189–1232. https://doi.org/10.1214/aos/1013203451
  9. Gentry, C., 2009. A Fully Homomorphic Encryption Scheme. Ph.D. Dissertation, Stanford University, Stanford, CA, USA.
  10. Genuer, R., Poggi, J.-M. and Tuleau-Malot, C., 2010. Variable selection using Random Forests. Pattern Recognition Letters, 31(14), 2225–2236. https://doi.org/10.1016/j.patrec.2010.03.014
  11. He, H. and Garcia, E.A., 2009. Learning from imbalanced data. IEEE Transactions on Knowledge and Data Engineering, 21(9), 1263–1284. https://doi.org/10.1109/TKDE.2008.239
  12. Hyndman, R.J. and Athanasopoulos, G., 2018. Forecasting: Principles and Practice. OTexts: Melbourne, Australia. OTexts.com/fpp2.
  13. James, G., Witten, D., Hastie, T. and Tibshirani, R., 2013. An Introduction to Statistical Learning. Vol. 112, Springer, New York.
  14. Kadhe, S.R. et al., Privacy-Preserving Federated Learning over Vertically and Horizontally Partitioned Data for Financial Anomaly Detection, arXiv:2310.19304. https://doi.org/10.48550/arXiv.2310.19304
  15. Ke, G. et al., 2017. LightGBM: A highly efficient gradient boosting decision tree. In Advances in Neural Information Processing Systems, Long Beach, CA, USA. pp. 3146–3154.
  16. Lee, J.-W. et al., 2022. Privacy-preserving machine learning with fully homomorphic encryption for deep neural network. IEEE Access, 10, 30039–30054. https://doi.org/10.1109/ACCESS.2022.3159694
  17. Legiest, W., Turan, F., Van Beirendonck, M., D'Anvers, J. P., & Verbauwhede, I., 2023. Neural network quantisation for faster homomorphic encryption. 29th International Symposium on On-Line Testing and Robust System Design (IOLTS), 1–3.
  18. Liaw, A. and Wiener, M., 2002. Classification and Regression by Random Forest. R News, 2(3), 18–22.
  19. Marcolla, C., Sucasas, V., Manzano, M., Bassoli, R., Fitzek, F. H., & Aaraj, N. 2022. Survey on fully homomorphic encryption, theory, and applications. Proceedings of the IEEE, 110(10), 1572–1609. https://doi.org/10.1109/JPROC.2022.3205665
  20. Miles, J., 2005. R‐squared, adjusted R‐squared. Encyclopedia of Statistics in Behavioral Science. https://doi.org/10.1002/0470013192.bsa526
  21. Naresh, V.S. and Ayyappa, D., 2024. PPDNN-CRP: Privacy-preserving deep neural network processing for credit risk prediction in cloud: A homomorphic encryption-based approach. Journal of Cloud Computing, 13, 149. https://doi.org/10.1186/s13677-024-00711-y
  22. Pattnaik, D., Ray, S., & Raman, R. (2024). Applications of artificial intelligence and machine learning in the financial services industry: A bibliometric review. Heliyon, 10(1), e23492. https://doi.org/10.1016/j.heliyon.2023.e23492
  23. Shah, S.B., 2025. Evaluating the effectiveness of machine learning in forecasting financial market trends: A fintech perspective. 3rd International Conference on Integrated Circuits and Communication Systems (ICICACS), Raichur, India, pp. 1-6.
  24. Yeom, S., Giacomelli, I., Fredrikson, M. and Jha, S., 2018. Privacy risk in machine learning: Analyzing the connection to overfitting. In Proceedings of the 2018 IEEE 31st Computer Security Foundations Symposium (CSF), 268–282.
  25. Willmott, C.J. and Matsuura, K., 2005. Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance. Climate Research, 30(1), 79–82.
  26. Xu, R., Baracaldo, N. and Joshi, J., 2021, Privacy-preserving machine learning: Methods, challenges and directions, arXiv:2108.04417. https://doi.org/10.48550/arXiv.2108.04417
  27. Shinde, S., 2025. What Are Machine Learning Applications? Top 10 Industry and Real-world Use Cases. Emeritus blog. Available at: https://emeritus.org/blog/machine-learning-what-are-machine-learning-applications, (28.04.2025)
  28. Yilmaz, E., Akturk, B., Civelek, D. and Buyuktanir, T., 2025. Audit opinions of Turkish Public Companies [Dataset]. Kaggle. Available at: https://www.kaggle.com/datasets/agrafintech/financial-data-of-turkish-public-companies
  29. Zama, 2023. Concrete ML: Privacy-preserving machine learning using fully homomorphic encryption. GitHub repository. Available at: https://github.com/zama-ai/concrete-ml