Intelligent Action-Based Fraud Detection System for Monitoring Emissions Testing in Motor Vehicle Inspection

In an effort to reduce road crash-related accidents and deaths in the Philippines from mechanical defects, the motor vehicle inspection system (MVIS) was mandated to assess the compliance of a vehicle in terms of its road-worthiness and safety emission standards. However, inspection processes are susceptible to possible fraud or human error, which can jeopardize the integrity of the system. To address this, fraud detection mechanisms have been explored in various domains, including healthcare, but their application in MVIS remains largely unexplored. As such, this study aims to develop an intelligent system for MVIS to validate inspection procedures. The approach involves developing a wireless body area network of sensors worn by the vehicle inspector, which collects inertial measurement unit data from different parts of the body. This data is sent to a convolutional neural network-based model, which predicts the actions of the inspector and whether the inspection process is fraudulent or not. Among all the various MVIS tests, the emissions testing procedure was chosen for this study because it requires the inspector to physically insert a probe into the exhaust pipe, a step that is vulnerable to human error and has been previously linked with cases of fraudulent results. The resulting hardware design is validated through test results demonstrating reliable data transmission with minimal packet loss. Additionally, the model achieves an accuracy of 96% in predicting whether the test procedure was fraudulent. Overall, this study offers valuable insights into the development of a wireless body area network of wearable devices capable of recognizing actions to detect fraudulent activity, specifically within the context of MVIS emissions testing procedures.