- Privacy and IP protection are increasingly important aspects of AI applications. Developing and optimizing safety critical automated driving systems requires storing, sharing, and processing of privacysensitive data for offline use. This includes, e.g., front-video camera images and videos that are recorded by a test fleet. These could contain Personal Identifiable Information (PII) such as faces or license plates. The European General Data Protection Regulation (GDPR) and similar privacy regulations in other countries set boundaries to the usage, storage, and sharing of these data. In this work, we are using Trusted Execution Environments (TEEs) as a Privacy Enhancing Technology (PET) to allow confidential end-to-end training on data that contains PII with drastically reduced legal risks under data protection regulations. We present a novel secure and scalable proof-of-concept using cloud-native technologies.
