Privacy-preserving Signal Detection, Analysis, and Classification in Automotive and Industrial Applications

• Applicant: Professor Dr.-Ing. Rainer Martin
• Subject Area: Communication Technology and Networks, High-Frequency Technology and Photonic Systems, Signal Processing and Machine Learning for Information Technology
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 549576906

Sensor networks are commonly used in various practical applications, such as smart homes, automotive, and large-scale industrial production processes. These networks often employ a client-server processing model due to the distributed nature of sensors and computational resources. However, without additional measures, sensor networks may disclose sensitive information about local conditions, including real-time information about industrial manufacturing processes or user-related data in an autonomous vehicle such as conversational speech. The transmission of data-rich signal representations from distributed sensors, in conjunction with elaborate deep learning algorithms in the cloud, can pose serious privacy risks. It is therefore important to consider these risks when designing such systems. In this transfer project, we expand our previous work on privacy-by-design feature extraction methods and apply them to new use cases in the automotive and industrial sectors. We utilize a client-server approach to transmit locally collected data through an information bottleneck to a central server for anomaly detection and condition monitoring. The objective of this study is to explore how privacy-preserving methods, previously developed and evaluated in basic experimental settings, can be adapted and utilized in more complex real-life applications. Our aim is to extend these approaches to cope with heterogeneous data distributions and domain shifts. Two application partners, one in industrial engineering and one in automotive, will contribute real-life data collection, evaluation efforts, and demonstration platforms. We expect to make significant progress towards a practical application of information bottlenecks and learning in distributed client-server systems.

• DFG Programme: Research Grants (Transfer Project)
• Application Partner: Auto Intern GmbH; VW Infotainment GmbH