Secure and Explainable Fraud Detection in Finance via Hierarchical Multi-source Dataset Distillation

2025-12-26

AI Summary

To address privacy leakage and lack of interpretability in collaborative financial fraud detection, this paper proposes a privacy-preserving and interpretable synthetic data generation framework. Our method introduces “tree-structured region distillation”: it converts a random forest into an axis-aligned hyperrectangular rule space and uniformly samples within it to generate a compact synthetic dataset—preserving local feature interactions while achieving full anonymization of original records. The framework simultaneously defends against membership inference attacks (attack accuracy ≈ 0.5), ensures global rule auditability, enables instance-level explanation generation, and provides calibration-aware uncertainty quantification. Evaluated on the IEEE-CIS dataset, the synthetic data achieves 85–93% compression, maintains stable AUC scores of 0.641–0.645, and boosts cross-cluster AUC to 0.687 after inter-institutional sharing.

Technology Category

• Machine Learning: Privacy
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Security and Privacy: Data transparency and provenance
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems

Abstract

We propose an explainable, privacy-preserving dataset distillation framework for collaborative financial fraud detection. A trained random forest is converted into transparent, axis-aligned rule regions (leaf hyperrectangles), and synthetic transactions are generated by uniformly sampling within each region. This produces a compact, auditable surrogate dataset that preserves local feature interactions without exposing sensitive original records. The rule regions also support explainability: aggregated rule statistics (for example, support and lift) describe global patterns, while assigning each case to its generating region gives concise human-readable rationales and calibrated uncertainty based on tree-vote disagreement. On the IEEE-CIS fraud dataset (590k transactions across three institution-like clusters), distilled datasets reduce data volume by 85% to 93% (often under 15% of the original) while maintaining competitive precision and micro-F1, with only a modest AUC drop. Sharing and augmenting with synthesized data across institutions improves cross-cluster precision, recall, and AUC. Real vs. synthesized structure remains highly similar (over 93% by nearest-neighbor cosine analysis). Membership-inference attacks perform at chance level (about 0.50) when distinguishing training from hold-out records, suggesting low memorization risk. Removing high-uncertainty synthetic points using disagreement scores further boosts AUC (up to 0.687) and improves calibration. Sensitivity tests show weak dependence on the distillation ratio (AUC about 0.641 to 0.645 from 6% to 60%). Overall, tree-region distillation enables trustworthy, deployable fraud analytics with interpretable global rules, per-case rationales with quantified uncertainty, and strong privacy properties suitable for multi-institution settings and regulatory audit.

Problem

• Develop a privacy-preserving dataset distillation method for financial fraud detection
• Generate explainable synthetic data while maintaining model performance and privacy
• Enable collaborative fraud analytics across institutions with auditable and interpretable results

Innovation

• Tree-region distillation creates compact, auditable synthetic datasets
• Uniform sampling within rule hyperrectangles preserves local feature interactions
• Explainability via aggregated rule statistics and per-case rationales