- Cet évènement est passé
AI4EU Web Café : « Towards robust and privacy-preserving AI » (15h00)
29 avril 2020 de 15 h 00 min
On April 29th, 2020, at 3:00 pm
Dr. Mathieu Sinn, IBM Research (Ireland) will present the live AI4EU Web Café Session themed :
« Towards robust and privacy-preserving AI »
Description of his live presentation :
In this talk I will address recent work towards robust and privacy-preserving AI. Without any doubt, AI has tremendous potential to disrupt existing businesses, create innovation and help solving most complex societal and environmental problems. However, it also has been demonstrated that AI is vulnerable to different sorts of adversarial attacks, which puts at risk its deployment in mission-critical applications. Examples of such attacks are poisoning of training data, e.g. in order to introduce backdoors, or evasion attacks at test time to compromise the integrity of AI models. Another potential problem is the leaking of proprietary or personal information through the process of training AI models. I will show a demonstration of those threats on actual state-of-the-art image and text classifiers. As one of the efforts to address those threats, I will present two open source projects that have been created by my team: The Adversarial Robustness 360 Toolbox (https://github.com/IBM/adversarial-robustness-toolbox) and the Differential Privacy Library (https://github.com/IBM/differential-privacy-library). Finally, I will talk about the H2020 MUSKETEER project which aims at providing robust and privacy-preserving AI via a Federated Machine Learning approach, with real-world demonstrations on use cases from the Manufacturing/Automotive and Healthcare industries.
Dr. Mathieu Sinn is a Research Staff Member and Manager of the AI, Security & Privacy group at the IBM Research lab in Dublin, Ireland. He has a Master’s in Computer Science and a PhD in Mathematics from the University of Lubeck, Germany. He has worked on a large variety of fundamental and practical aspects of Machine Learning, with a recent focus on robustness of AI against adversarial threats. Mathieu is Data Science Thought Leader certified by The Open Group, regular reviewer for top AI conferences and has served as external PhD committee member on various occasions.