[Back]

C. Lachner, T. Rausch, S. Dustdar:
"A Privacy Preserving System for AI-assisted Video Analytics";
Talk: 5th IEEE International Conference on Fog and Edge Computing (ICFEC 2021) - Online Conference, Melbourne, Australia; 2021-05-10; in: "Proceedings of the 5th IEEE International Conference on Fog and Edge Computing (ICFEC 2021)", Y. Simmhan, B. Varghese, L. Mashayckhy, R. Buyya, O. Rana (ed.); IEEE, (2021), ISBN: 978-1-6654-0292-7; 74 - 78.

The emerging Edge computing paradigm facilitates the deployment of distributed AI-applications and hardware, capable of processing video data in real time. AI-assisted video analytics can provide valuable information and benefits for parties in various domains. Face recognition, object detection, or movement tracing are prominent examples enabled by this technology. However, the widespread deployment of such mechanism in public areas are a growing cause of privacy and security concerns. Data protection strategies need to be appropriately designed and correctly implemented in order to mitigate the associated risks. Most existing approaches focus on privacy and security related operations of the video stream itself or protecting its transmission. In this paper, we propose a privacy preserving system for AI-assisted video analytics, that extracts relevant information from video data and governs the secure access to that information. The system ensures that applications leveraging extracted data have no access to the video stream. An attribute-based authorization scheme allows applications to only query a predefined subset of extracted data. We demonstrate the feasibility of our approach by evaluating an application motivated by the recent COVID-19 pandemic, deployed on typical edge computing infrastructure.

privacy, artificial-intelligence, edge-computing, information-extraction, video-processing, attribute based authentication

http://dx.doi.org/10.1109/ICFEC51620.2021.00018

Project Head Schahram Dustdar:
FogProtect