Graph-Based Fraud Detection for Mobile Ad Install Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing fraud detection systems in mobile applications are ineffective in identifying and preventing click fraud and transaction fraud, where publishers use bots to generate fake clicks and installs, leading to a loss of advertisers' marketing budgets.
Innovation Solution
A computer system that collects device and application data, generates graphs based on user behavior and parameters like time between events, analyzes these graphs with trained data to score publishers, and blocks them if their score exceeds a predefined level, using a signal generator circuitry to trigger hardware components for real-time fraud detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing fraud detection systems are used, then publishers can operate without restriction, but advertisers suffer budget loss from fake traffic and installs
Solution Approach 1:
The system performs preliminary fraud detection by analyzing device data, application data, and behavior patterns before advertisers spend their budgets on fraudulent traffic. The fraud score is calculated in advance based on multiple parameters including time between events, hardware component triggering patterns, and user behavior deviations, allowing preventive blocking of fraudulent publishers before budget loss occurs
Solution Approach 2:
The system implements continuous feedback loops where fraud detection results are fed back to adjust detection thresholds and parameters. The fraud score calculation incorporates real-time analysis of device data, application data, and behavior patterns, with the ability to dynamically adjust detection sensitivity based on emerging fraud patterns, improving detection accuracy while reducing false positives
2Productivity
If real-time fraud detection and blocking is implemented, then fraudulent publishers are prevented, but system complexity increases
Solution Approach 1:
The fraud detection system is segmented into distinct modular components: device data collection module, application data collection module, behavior pattern analysis module, fraud score calculation module, and publisher blocking module. Each module handles specific aspects of fraud detection independently, making the complex system manageable and maintainable while enabling real-time operation through parallel processing
Solution Approach 2:
The system dynamically adjusts detection parameters such as time thresholds between events, fraud score thresholds for blocking, and sensitivity levels based on learned fraud patterns. This allows the system to adapt to different types of fraud without requiring complete system redesign, managing complexity through flexible parameter adjustment rather than structural complexity
Data Source
AI summary
The present disclosure provides a method and system to detect advertisement fraud. The fraud detection platform receives device data and application data associated with one or more advertisements published on at least one publisher on one or more media devices. In addition, the fraud detection platform identifies a plurality of parameters based on the device data and the application data. Further, the fraud detection platform generates a plurality of graphs based on the device data, the application data, user behavior and the plurality of parameters. Furthermore, the fraud detection platform analyzes the plurality of graphs with trained data to identify fraud based on the deviation.