On-Device Behavioral Biometrics for Mobile Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fraud prevention techniques for mobile computing devices are inadequate in detecting unauthorized access, especially when identification information is stolen, as they rely on biometric identification, passwords, and security keys, which can be compromised.
Innovation Solution
A method and system utilizing machine learning to create and compare behavioral profiles on mobile computing devices, tracking navigation and identification information to detect anomalies and prevent unauthorized access to sensitive data, by implementing a local machine learning system that reduces latency and processing load on cloud systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If biometric identification, passwords, and security keys are used for fraud prevention, then identification security is improved, but the system becomes vulnerable when identification information is stolen
Solution Approach 1:
The patent introduces behavioral biometrics as an intermediary layer between traditional identification methods and access control. Instead of relying solely on static identification information (passwords, biometric data), the system uses behavioral patterns (typing rhythm, navigation patterns, device interaction) as a mediator to verify user identity continuously, making stolen identification information insufficient for unauthorized access
Solution Approach 2:
The system transitions from static identification parameters (fixed passwords, unchanging biometric templates) to dynamic behavioral parameters that change with each user interaction. By monitoring and analyzing varying behavioral characteristics over time, the system creates a moving target that cannot be compromised by stealing a single set of identification credentials
2Measurement precision
If cloud-based machine learning systems are used for anomaly detection, then detection accuracy is improved, but processing latency and cloud system load increase
Solution Approach 1:
The patent segments the anomaly detection system into two parts: a lightweight local model on the mobile device for real-time inference, and a cloud-based system for periodic model retraining and updates. This segmentation allows immediate local detection (low latency) while maintaining high accuracy through cloud-based model improvements
Solution Approach 2:
The system performs preliminary action by pre-training machine learning models in the cloud and deploying them to mobile devices. The local models are pre-configured with behavioral profiles and detection algorithms, enabling immediate anomaly detection without real-time cloud processing, thus reducing latency while maintaining detection capabilities
3Reliability
If continuous behavioral tracking is implemented, then fraud detection capability is improved, but user privacy and data security concerns increase
Solution Approach 1:
The patent applies local quality by processing and analyzing behavioral data locally on the user's mobile device rather than transmitting raw data to centralized servers. The machine learning model operates on-device, converting raw behavioral inputs into anonymized insights locally, thus maintaining fraud detection capability while minimizing privacy exposure through localized data handling
Data Source
AI summary
A system and method for detecting an anomaly based on user interaction with a mobile computing device is disclosed. The method includes activating, on the mobile computing device, an application, which is configured to store the machine learning model and the behavioral profile generated by the machine learning model in the memory of the mobile computing device; receive, track, and store in the memory of the mobile computing device an input pattern including navigation information and identification information inputted from the user during the user interaction with the mobile computing device; detect anomaly based on the user interaction with the mobile computing device by comparing the stored behavioral profile with the stored input pattern; and prohibit user to have further access to the financial application including user accounts in response to detection of the anomaly.