SeSaMe Seminar by Dr Thomas Winkler
Title: Protecting and Assessing Privacy in Visual Data
Speaker: Dr Thomas Winkler
Date & Time: 14th April 2015 (Tue), 2pm - 3pm
Venue: Seminar Room, Level 1, iCube Building
Abstract: Visual sensing is becoming more and more widespread not only for public but also private monitoring applications. Especially in private environments, privacy protection is a critical issue. This talk discusses approaches to bring privacy protection close to the data source and to make it an integral part of the sensing device itself. Privacy protection on an embedded camera device comes with many challenges including limited system resource as well as limited reliability for the identification of privacy sensitive image regions. In this talk, a cartooning technique is presented which is optimized for low-performance hardware and, by applying it globally, eliminates the need for region of interest detection. The objective and automated evaluation of the effectiveness of privacy protection techniques is not yet commonplace within the community. This talk presents a novel evaluation framework which allows assessment of the level of privacy protection and the remaining utility of the filtered data. The talk also briefly covers novel techniques for platform security which are crucial for any privacy-preserving system.
Bio: Thomas Winkler studied Telematics at Graz University of Technology and earned his Dipl. Ing. (MSc) with distinction in 2005. He then joined the Institute for Applied Information Processing and Communications (IAIK) as a researcher in the EU project "Open Trusted Computing" where he designed the first Java software stack for "Trusted Computing". In May 2007, he joined the Institute of Networked and Embedded Systems (NES) at Alpen-Adria-Universität Klagenfurt. His research interests focus on embedded smart camera systems including both hardware and software. Security and privacy in visual sensor networks have become central topics of his work. In July 2011 he completed his PhD studies with distinction. As part of his research, he developed a TPM-enabled embedded camera system called TrustCAM with inherent security and privacy protection features. In his current work, in the TrustEYE project (http://trusteye.aau.at), he advances these ideas and develops concepts for a secure sensing unit. A key aspect is the sensor-level separation of non-sensitive from sensitive data. In a parallel project called ProSecCo, he explores the applicability of Physical Unclonable Functions (PUFs) to secure embedded sensing devices.