A Dual-Path Generative Framework for Zero-Day Fraud Detection in Banking Systems

2026-03-17 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy, Data Science & Analytics · Depth: Advanced, long

Summary

This paper introduces a Dual-Path Generative Framework for zero-day fraud detection in high-frequency banking systems, addressing the critical trade-off between low-latency detection and GDPR-mandated explainability. The architecture employs a Variational Autoencoder (VAE) for real-time anomaly detection based on reconstruction error, ensuring sub-50ms inference latency, while an asynchronous Wasserstein GAN with Gradient Penalty (WGAN-GP) synthesizes high-entropy fraudulent scenarios to mitigate extreme class imbalance and stress-test detection boundaries. To handle discrete banking data like Merchant Category Codes, a Gumbel-Softmax estimator is integrated, and a trigger-based SHAP (Shapley Additive Explanations) mechanism is activated only for high-uncertainty transactions, reconciling XAI computational cost with real-time throughput. This framework effectively counters adaptive micro-transaction fraud, Card-Not-Present (CNP) velocity attacks, and Account Takeover (ATO) scenarios, ensuring both operational resilience and regulatory compliance through a human-in-the-loop feedback mechanism.

Key takeaway

A Dual-Path Generative Framework tackles zero-day fraud in banking by coupling a VAE for real-time anomaly detection with an asynchronous WGAN-GP for adversarial synthesis. It achieves <50ms inference latency and GDPR compliance by integrating Gumbel-Softmax for discrete data and selectively triggering SHAP explainability. This robustly mitigates extreme class imbalance and adapts to evolving fraud patterns.

Topics

• Zero-Day Fraud Detection
• Generative Adversarial Networks
• Variational Autoencoders
• Explainable AI
• Banking Systems

Best for: AI Scientist, Research Scientist, AI Researcher, Machine Learning Engineer, AI Security Engineer

Related on AIssential

• Synthetic data in cryptocurrencies using generative models — CGANs with LSTM generators can effectively synthesize cryptocurrency time series, preserving market dynamics.
• UniGenDet: A Unified Generative-Discriminative Framework for Co-Evolutionary Image Generation and Generated Image Detection — UniGenDet unifies image generation and detection through symbiotic mechanisms for co-evolutionary performance enhancement.
• A Fraudster’s Paradise — Generative AI significantly amplifies the scale and sophistication of online fraud, impacting both professional criminals and casual cheaters.
• Generative Adversarial Networks (GANs) - Explained — GANs use adversarial competition between two neural networks to learn and generate data indistinguishable from real samples.
• Unlocking the relational power of graph neural networks in fraud prevention — Graph Neural Networks enhance fraud detection by modeling relational data, uncovering networked fraud invisible to traditional tabular methods.
• Sun Finance automates ID extraction and fraud detection with generative AI on AWS — Combining specialized OCR with LLM structuring significantly enhances ID extraction accuracy and efficiency.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.