WIP: Blurred Lines -- A GDPR-Compliant Framework for Anonymising Automotive Video Data

Protecting personally identifiable information (PII) is now a mandatory requirement, in the automotive domain, as this information can be used to potentially identify, track, or exploit an individual. The General Data Protection Regulation (GDPR) ensures this information is protected by ensuring that organisations dealing with such details follow stringent rules. This is especially pertinent to video-feed data which is collected by the vehicle; an integral feature of most modern vehicles. This paper presents a framework that ensures that the video data obtained from and used to train autonomous vehicles is GDPR-compliant and can be easily shared among companies by ensuring that PII such as faces and number-plates are masked, and can be shared legally. This framework proposes five methods to identify the PII and various approaches to blur them. This work was done alongside an industrial partner, that provided a real-world dataset to test the framework's accuracy in different scenarios. Results of our experiments ensure near 100% accuracy in detecting and blurring PIIs. The framework also has an easy-to-use command-line interface and can be easily modified to include more methods in the future. The tool has already been adapted for use in industry.