Federated Learning Framework for Privacy-Preserving Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for processing digital transactions face challenges in real-time processing, security, and fraud detection due to high transaction volumes, siloed data systems, and the need for accurate machine learning models that can handle large-scale, decentralized data while preserving privacy.
Innovation Solution
A federated learning framework for anomaly detection in financial transactions that uses private differentiable models, attention mechanisms, and differential privacy techniques to share privacy-preserving outputs across institutions, avoiding model aggregation and pooling, and employing k-anonymity and resampling to protect data privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional centralized machine learning models are used for fraud detection, then detection accuracy can be improved, but data privacy and security are compromised due to the need to pool sensitive transaction data from multiple institutions
Solution Approach 1:
The patent segments the centralized machine learning system into multiple federated learning nodes, each representing a different financial institution. Each node trains locally on its own data without sharing raw transaction records, thus maintaining data privacy while collectively improving fraud detection accuracy through model aggregation. The segmentation allows each institution to retain control over its sensitive data while contributing to a more accurate global model.
Solution Approach 2:
The patent introduces a central server as an intermediary that coordinates the federated learning process without accessing sensitive institutional data. This mediator facilitates model aggregation and distribution of update rules while maintaining data isolation. The intermediary enables collaboration between institutions without requiring direct data sharing, thus improving detection accuracy while preserving privacy.
2Measurement precision
If more transaction data is collected and analyzed for fraud detection, then detection accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The patent implements preliminary action by pre-training fraud detection models on historical transaction data before deployment. The federated learning system performs offline model aggregation and update rule distribution, so that when real-time transaction analysis is needed, the models are already optimized and ready for rapid inference. This preliminary preparation reduces processing time during actual fraud detection operations.
3Reliability
If machine learning models are trained on decentralized data from multiple institutions, then fraud detection capability improves, but system complexity increases due to coordination and privacy preservation requirements
Solution Approach 1:
The patent implements a universal federated learning framework that can be adopted by any financial institution regardless of their specific data characteristics or infrastructure. The system uses standardized model architectures and communication protocols that work across different institutions, reducing the complexity of integration. The multi-functional design allows the same framework to handle various types of financial transactions and fraud patterns while maintaining a consistent approach to privacy preservation and model aggregation.
Data Source
AI summary
Various measures to preserve privacy and to train neural network models are provided. A federated learning system comprises a payment decisioning server associated with a payment decisioning entity and a financial institution server associated with a financial institution. A transaction involves a customer of the financial institution. In an example method, the payment decisioning server transmits, to the financial institution server, a query relating to the customer. In the example method, the payment decisioning server receives, from the financial institution server, a response to the query, the response comprising an account embedding, the account embedding having been derived by the financial institution server using account information associated with the customer as input to a financial institution private embedding model. In the example method, the payment decisioning server generates a transaction decision for the transaction using the account embedding and transaction data associated with the transaction.