Trustworthy Visual Computing: Towards Sustainable Detection of Deep Fakes
Event date:  November 30, 2020, 4:00 PM  to 6:00 PM

Abstract 

With the advance of deep learning techniques, large scale computation and large datasets, data-driven image synthesis techniques have achieved a qualitative new level of performance in recent years. Most notably Generative Adversarial Networks (GANs) allow the synthesis of visual data that can already now be indistinguishable to real visual data from a human observer point of view. While this opens up new applications, this has several serious consequences on trust in visual data and can even be used to manipulate public opinions. Hence, we seek understanding in which ways such technology can be used beyond the intended use case in order to inform the model creators about potential dual use scenario. Furthermore, we investigate possibilities to detect such deep fakes. Unfortunately, this often leads to an arms race which does not allow for a sustainable detection of such fake images. Therefore, our most recent research investigates different watermarking and fingerprinting techniques that support a responsible disclosure of such state of the art models in order to mitigate potential risks they pose.


About the Speaker

Mario Fritz is faculty member at the CISPA Helmholtz Center for Information Security, professor at the Saarland University, and founding member of the ELLIS unit Saarbruecken. Before, he was senior researcher and research group head at the Max Planck Institute for Informatics, and PostDoc at the International Computer Science Institute and UC Berkeley. He did his PhD at the TU Darmstadt. His current work is centered around Trustworthy Information Processing with a focus on the intersection of AI & Machine Learning with Security & Privacy. He served as Area Chair for major computer vision conferences (ECCV, ICCV), associate editor of IEEE TPAMI, and is a member of the ACM Europe Technical Policy Committee Europe and working group "AG Nutzbarmachung digitaler Daten für KI-Entwicklungen in der Gesundheitsforschung" in the "Forum Gesundheit". He is also a leading scientist of the Helmholtz Medical Security, Privacy, and AI Research Center, where he is coordinating projects on trustworthy federated data analytics and protecting genetic data with synthetic cohorts from deep generative models. He has co-organized the workshop "The Bright and the Dark Side of Computer Vision: Challenges and Opportunities for Privacy and Security'" from 2018 to 2020.


About the Series

The Lecture Series is organized every winter term and consist of talks with international speakers. The speakers are experts from various fields and professions. The talks are aimed at covering the whole spectrum of visual computing and at discussing the relevance of quantification. This years lecture series will be held entirely online.


This is an online event only.

Registration: If you would like to attend, please email Leonel Merino to register.