Generative AI for Credit Risk Management: Risk-Conditional Diffusion with Holdout-Calibrated Release Gates for Empirically Privacy-Audited Synthetic Credit Data
DOI:
https://doi.org/10.54691/dgpq5t90Keywords:
Generative AI, credit risk, synthetic data, diffusion models, membership inference, empirical privacy auditing, model risk managementAbstract
Generative AI is increasingly used to share and augment confidential credit data, but high-fidelity generators can reveal which borrowers they were trained on. We study this tension on three public credit risk datasets and propose PG-Diff (proximity-gated diffusion), which couples a risk-conditional tabular diffusion model, combining mass-point encoding of zero-inflated amounts with classifier-free guidance on the default label, with a generator-agnostic release gate. The gate enforces absolute proximity quotas and a relative proximity balance, both calibrated on held-out real records so that records drawn independently of the training data pass them in expectation. Adopting the fidelity, utility and privacy evaluation framework of Yang and Yang, extended with low-false-positive membership inference and a model-risk agreement metric, we compare SMOTE, Gaussian copula, TVAE, CTGAN and diffusion releases over up to three random splits. Conditional diffusion reaches train-on-synthetic AUCs of 0.751, 0.845 and 0.748 on Taiwan, Give Me Some Credit and German (real-data models: 0.762, 0.834, 0.778), gives the most realistic records among the deep generators, and shows membership-inference AUCs close to 0.5 on the two larger datasets. Interpolation-based SMOTE attains similar utility but leaks heavily (attack AUC up to 0.76, true-positive rate at 1% false positives up to 0.58); the gate lowers that rate by 43% on average and flags SMOTE through relaxation rates of up to 97%. For non-interpolating generators the gate costs at most 0.039 AUC, and in a memorization stress test it moves the DCR share from 0.595 to 0.472. Holdout-calibrated release gates thus provide a generator-agnostic empirical safeguard that complements formal mechanisms such as differential privacy.
Downloads
References
[1] Assefa, S. A., Dervovic, D., Mahfouz, M., Tillman, R. E., Reddy, P., & Veloso, M. (2020). Generating synthetic data in finance: Opportunities, challenges and pitfalls. In Proceedings of the 1st ACM International Conference on AI in Finance (ICAIF). https://doi.org/10.1145/3383455.3422554 DOI: https://doi.org/10.1145/3383455.3422554
[2] Wiese, M., Knobloch, R., Korn, R., & Kretschmer, P. (2020). Quant GANs: Deep generation of financial time series. Quantitative Finance, 20(9), 1419–1440. https://doi.org/10.1080/14697688.2020.1730426 DOI: https://doi.org/10.1080/14697688.2020.1730426
[3] Wu, S., Irsoy, O., Lu, S., Dabravolski, V., Dredze, M., Gehrmann, S., Kambadur, P., Rosenberg, D., & Mann, G. (2023). BloombergGPT: A large language model for finance. arXiv:2303.17564.
[4] Yang, Y., & Yang, J. (2026). Synthetic data meets finance: Generative models for privacy preserving analytics. Journal of Banking Financial Dynamics, 10(4), 1–8. https://doi.org/10.55220/2576-6821.v10.928 DOI: https://doi.org/10.55220/2576-6821.v10.928
[5] Stadler, T., Oprisanu, B., & Troncoso, C. (2022). Synthetic data—Anonymisation groundhog day. In Proceedings of the 31st USENIX Security Symposium (pp. 1451–1468).
[6] Kotelnikov, A., Baranchuk, D., Rubachev, I., & Babenko, A. (2023). TabDDPM: Modelling tabular data with diffusion models. In Proceedings of the 40th International Conference on Machine Learning (ICML) (Vol. 202, pp. 17564–17579). PMLR.
[7] Zhang, H., Zhang, J., Srinivasan, B., Shen, Z., Qin, X., Faloutsos, C., Rangwala, H., & Karypis, G. (2024). Mixed-type tabular data synthesis with score-based diffusion in latent space. In Proceedings of the International Conference on Learning Representations (ICLR).
[8] Carlini, N., Hayes, J., Nasr, M., Jagielski, M., Sehwag, V., Tramèr, F., Balle, B., Ippolito, D., & Wallace, E. (2023). Extracting training data from diffusion models. In Proceedings of the 32nd USENIX Security Symposium (pp. 5253–5270).
[9] Chen, J., Zhang, X., & Han, X. (2023). Explainable identification of tax-related financial reporting anomalies in U.S. public companies: Cross-fitted tax-consistency residuals and Owen-value Shapley propagation. World Journal of Economics and Business Research, 1(1), 33–45. DOI: https://doi.org/10.61784/wjebr4125
[10] Board of Governors of the Federal Reserve System & Office of the Comptroller of the Currency. (2011, April). Supervisory guidance on model risk management (SR Letter 11-7).
[11] Dwork, C., & Roth, A. (2014). The algorithmic foundations of differential privacy. Foundations and Trends in Theoretical Computer Science, 9(3–4), 211–407. https://doi.org/10.1561/0400000042 DOI: https://doi.org/10.1561/0400000042
[12] 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 ACM SIGSAC Conference on Computer and Communications Security (CCS) (pp. 308–318). https://doi.org/10.1145/2976749.2978318 DOI: https://doi.org/10.1145/2976749.2978318
[13] Jordon, J., Yoon, J., & van der Schaar, M. (2019). PATE-GAN: Generating synthetic data with differential privacy guarantees. In Proceedings of the International Conference on Learning Representations (ICLR).
[14] Zuo, M., Kang, I., Patterson, S., & Seneviratne, O. (2026). Measuring privacy risks and tradeoffs in financial synthetic data generation. In Companion Proceedings of the ACM Web Conference (WWW Companion). https://doi.org/10.1145/3774905.3797244 DOI: https://doi.org/10.1145/3774905.3797244
[15] Ganev, G., & De Cristofaro, E. (2025). The inadequacy of similarity-based privacy metrics: Privacy attacks against “truly anonymous” synthetic datasets. In Proceedings of the 46th IEEE Symposium on Security and Privacy (SP). DOI: https://doi.org/10.1109/SP61157.2025.00218
[16] Platzer, M., & Reutterer, T. (2021). Holdout-based empirical assessment of mixed-type synthetic data. Frontiers in Big Data, 4, 679939. https://doi.org/10.3389/fdata.2021.679939 DOI: https://doi.org/10.3389/fdata.2021.679939
[17] Ho, J., & Salimans, T. (2022). Classifier-free diffusion guidance. arXiv:2207.12598.
[18] Carlini, N., Chien, S., Nasr, M., Song, S., Terzis, A., & Tramèr, F. (2022). Membership inference attacks from first principles. In Proceedings of the IEEE Symposium on Security and Privacy (SP) (pp. 1897–1914). https://doi.org/10.1109/SP46214.2022.9833649 DOI: https://doi.org/10.1109/SP46214.2022.9833649
[19] Yue, X., & Zhu, R. (2024). An explainable hybrid machine learning framework for cash-flow-conditioned multi-horizon liquidity risk early warning: An SME-oriented benchmark study. World Journal of Management Science, 2(1), 63–72. DOI: https://doi.org/10.61784/wms4120
[20] Xu, L., Skoularidou, M., Cuesta-Infante, A., & Veeramachaneni, K. (2019). Modeling tabular data using conditional GAN. In Advances in Neural Information Processing Systems (Vol. 32).
[21] Patki, N., Wedge, R., & Veeramachaneni, K. (2016). The synthetic data vault. In Proceedings of the IEEE International Conference on Data Science and Advanced Analytics (DSAA) (pp. 399–410). https://doi.org/10.1109/DSAA.2016.49 DOI: https://doi.org/10.1109/DSAA.2016.49
[22] Ho, J., Jain, A., & Abbeel, P. (2020). Denoising diffusion probabilistic models. In Advances in Neural Information Processing Systems (Vol. 33, pp. 6840–6851).
[23] Kynkäänniemi, T., Karras, T., Laine, S., Lehtinen, J., & Aila, T. (2019). Improved precision and recall metric for assessing generative models. In Advances in Neural Information Processing Systems (Vol. 32).
[24] Song, J., Meng, C., & Ermon, S. (2021). Denoising diffusion implicit models. In Proceedings of the International Conference on Learning Representations (ICLR).
[25] Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., & Liu, T.-Y. (2017). LightGBM: A highly efficient gradient boosting decision tree. In Advances in Neural Information Processing Systems (Vol. 30, pp. 3146–3154).
[26] Bussmann, N., Giudici, P., Marinelli, D., & Papenbrock, J. (2021). Explainable machine learning in credit risk management. Computational Economics, 57(1), 203–216. https://doi.org/10.1007/s10614-020-10042-0 DOI: https://doi.org/10.1007/s10614-020-10042-0
[27] Yeh, I.-C., & Lien, C.-H. (2009). The comparisons of data mining techniques for the predictive accuracy of probability of default of credit card clients. Expert Systems with Applications, 36(2), 2473–2480. https://doi.org/10.1016/j.eswa.2007.12.020 DOI: https://doi.org/10.1016/j.eswa.2007.12.020
[28] Teng, D. (2025). TRIAD: Deadline- and reliability-aware scheduling for post-earthquake structural assessment workflows. World Journal of Engineering Research, 3(6), 55–66. DOI: https://doi.org/10.61784/wjer4124
[29] Hofmann, H. (1994). Statlog (German Credit Data). UCI Machine Learning Repository. https://doi.org/10.24432/C5NC77
[30] Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002). SMOTE: Synthetic minority over-sampling technique. Journal of Artificial Intelligence Research, 16, 321–357. https://doi.org/10.1613/jair.953 DOI: https://doi.org/10.1613/jair.953
[31] Han, X., et al. (2025). Symmetry-Aware Credit Risk Modeling: A Deep Learning Framework Exploiting Financial Data Balance and Invariance. Symmetry, 17(3), 341. https://doi.org/10.3390/sym17030341 DOI: https://doi.org/10.3390/sym17030341
[32] Chen, J., et al. (2026). Temporal Transformer with Conditional Tabular GAN for Credit Card Fraud Detection: A Sequential Deep Learning Approach. Mathematics, 14, 1183. https://doi.org/10.3390/math14071183 DOI: https://doi.org/10.3390/math14071183
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Academic Journal of Finance and Accounting

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







