Cameras have found their way into many parts of our daily lives. While monitoring public places is still a common use case for camera systems, applications in private environments are emerging. The ever growing number of cameras raises important questions concerning security and privacy. “Our approach to tackle these questions is to protect all sensitive data before it leaves the camera,” explains Bernhard Rinner. “We exploit the available computing resources of modern camera systems for onboard privacy protection and data security and do not rely on pure software solutions. We have successfully demonstrated this approach on several prototypes.”
The key idea is to “protect” access to the image sensor and encapsulate dedicated security and privacy functionality in a TrustEYE—a secure sensing unit embedded on the smart camera. The TrustEYE has exclusive access to the image sensor’s raw data. It separates sensitive from non-sensitive data by applying dedicated image analysis and ensures that only non-sensitive data is made available to the camera host system. “In another prototype, we use modern hybrid ARM/FPGA system on chip solutions to provide security and high speed image analysis functions,” Ihtesham Haider points out. “We exploit inherent hardware properties in the form of physical unclonable functions to realize high levels of security without requiring additional specialized hardware for cryptographic functions.”
Privacy protection is achieved by intentionally distorting sensitive regions of the captured images. In a first approach we have developed so-called cartooning privacy filters which preserve privacy while ensuring a minimum reduction of the image fidelity and run onboard of our TrustEYE platform.
Another approach was devoted towards privacy protection in recreational videography from small drones that can capture bystanders who may be uncomfortable about appearing in those videos. We have developed a robust spatio-temporal hopping blur filter that protects privacy through de-identification of face regions. Our filter distorts a face region with secret parameters to be robust to naïve, parrot and reconstruction attacks. The distortion is minimal and adaptive to the resolution of the captured face: we select the smallest Gaussian kernel that reduces the face resolution below a certain threshold. This privacy filter is for on-board installation and produces temporally smooth and pleasant videos.
O. Sarwar, B. Rinner, and A. Cavallaro. A privacy-preserving filter for oblique face images based on adaptive hopping Gaussian mixtures. IEEE Access, 2019.
B. Rinner. Can we trust smart cameras. Computer, 2019.
O. Sarwar, A. Cavallaro, and B. Rinner. Temporally smooth privacy protected airborne videos. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2018), Madrid, Spain. October 2018.
A. Erdelyi, T. Winkler, and B. Rinner. Privacy Protection vs. Utility in Visual Data: An Objective Evaluation Framework. Multimedia Tools and Applications, Springer, 2018.
I. Haider and B. Rinner. Private Space Monitoring with SoC-based Smart Cameras. In Proc. IEEE International Conference on Mobile Ad-hoc and Sensor Systems (MASS 2017), Olando, FL, USA. October 2017.
I. Haider, M. Hoeberl, and B. Rinner. Trusted Sensors for Participatory Sensing and IoT Applications based on Physically Unclonable Functions. In Proc. International Workshop IoT Privacy, Trust, and Security, 2016.
T. Winkler and B. Rinner. Security and Privacy Protection in Visual Sensor Networks: A Survey. ACM Computing Surveys, 2014.
A. Erdelyi, T. Barat, P. Valet, T. Winkler, and B. Rinner. Adaptive Cartooning for Privacy Protection in Camera Networks. In Proc. IEEE International Conference Advanced Video and Signal-Based Surveillance, 2014.