The Bright and Dark Sides of Computer Vision and Machine Learning — Challenges and Opportunities for Privacy and Security

Bernt Schiele

Max Planck Institute, DE

Abstract

Computer Vision has been revolutionized by Machine Learning and in particular Deep Learning. For many problems which have been studied for decades, state-of-the art performance has dramatically improved by using artificial neural networks. At the same time, more and more people share significant amounts of visual data e.g. in social networks. And while it is clear that visual data contains privacy relevant information, it is less clear which privacy implications visual data dissemination in the age of deep learning driven computer vision has and if and how it might be possible to control the leakage of privacy relevant information. Another challenge is the robustness of these algorithms against adversarial attacks and what that might imply for the deployment e.g. in autonomous vehicles. Last, but not least, I will discuss new insights about reverse engineering deep neural networks as well as stealing the entire functionality of them cheaply. While we are clearly at the infancy of understanding privacy as well as security implications of deep neural networks, the talks aims to raise awareness as well as to motivate more researchers to address these important challenges.