METHOD AND SYSTEM FOR DETECTION OF ADVERTISEMENT FRAUD

原文（中文）

A computer-implemented method is disclosed for detecting advertisement fraud in real-time using a multi-modal analysis framework. The method involves receiving user data and user action data from a media device, including demographic information and sensor-based interaction metrics such as accelerometer, gyroscope, and touch sensor data. These inputs are processed by an advertisement fraud detection system equipped with a processor and hardware-run algorithms, including a multi-modal machine learning model that distinguishes human from non-human interactions. Fraudulent actions are identified by detecting deviations from predefined behavioral baselines enriched with campaign-level intelligence. The system further analyzes historical ad performance to identify downtime periods characterized by low human activity and high fraud probability. During these periods, fake advertisements are adaptively inserted based on contextual mismatches to confirm fraudulent behavior. Upon detection, notifications are dispatched to advertisers via multiple communication mediums.