Edge Device Machine Learning for Privacy-Preserving Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing transaction processing systems face challenges in detecting anomalies and preventing fraud due to privacy measures that prohibit the transmission of user device data for centralized machine learning model training, leading to increased loss and user frustration.
Innovation Solution
A hybrid approach is implemented where machine learning models are trained partially or fully on edge devices using federated learning, with private data obfuscated and processed locally, and fine-tuned at a central server, enhancing security and user experience through silent authentication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If user device data is transmitted to a central server for machine learning model training, then the system can detect anomalies and prevent fraud more effectively, but privacy measures prohibit this transmission leading to increased loss and reduced security
Solution Approach 1:
The patent segments the centralized machine learning training process into distributed edge training on user devices and centralized model aggregation. Each user device trains local models using private data without transmitting the data itself, while only model updates or gradients are sent to the server. This segmentation resolves the contradiction by enabling fraud detection model training while respecting privacy restrictions that prohibit data transmission.
Solution Approach 2:
The patent introduces an intermediary mechanism where encrypted model updates or gradient information serve as the mediator between user device data and the central server. Instead of directly transmitting sensitive user data, the system transmits processed model information that contains fraud detection patterns without revealing private user information. This intermediary approach enables effective fraud detection while maintaining privacy protection.
2Measurement precision
If machine learning models are trained centrally with all user data, then fraud detection accuracy improves, but transmission latency and data privacy risks increase
Solution Approach 1:
The patent implements preliminary action by pre-training base models centrally and distributing them to edge devices before actual fraud detection scenarios occur. User devices then perform local fine-tuning training using their private data, preparing detection models in advance without requiring real-time data transmission. This preliminary model preparation reduces transmission latency while maintaining high detection accuracy through local adaptation to user-specific patterns.
3Object-affected harmful factors
If private user data is kept local on edge devices, then privacy is maintained, but the system loses the ability to perform centralized analysis for fraud prevention
Solution Approach 1:
The patent implements feedback mechanisms where edge-trained models send aggregated learning signals and model updates back to the central server. The server aggregates these feedback signals from multiple users and uses them to improve the global fraud detection model, which is then redistributed to edge devices. This feedback loop enables centralized fraud prevention capability while maintaining local data privacy, as only processed model information rather than raw user data is transmitted.
Data Source
AI summary
Techniques are disclosed in which a computing device repeatedly trains, using a stream of user data received at the computing device, a baseline model to generate a device-trained model, wherein the baseline model is trained at the computing device without providing user data included in the stream to a server computer system. In some embodiments, the computing device inputs, to the device-trained model, a set of characteristics associated with a user request received from a user of the computing device, wherein the device-trained model outputs a score for the user request. In some embodiments, the computing device transmits, to the server computer system, the score for the user request, wherein the transmitting includes requesting a decision for the user request. In some embodiments, the computing device performs an action associated with the user request in response to receiving a decision for the user request from the server computer system.