This project is joint work together with my co-authors @lubux86@AlexanderViand @KuechlerNicolas and @anwarhit , and you can find the paper here: https://t.co/Wsm5wEtqEm
Grateful for the opportunity to present our work on the robustness of secure federated learning at @IEEESSP!
We provide new insights into why attacks are possible and why a definitive solution to FL robustness is challenging. 1/
We show that certain classes of severe attacks can be mitigated effectively by enforcing constraints such as norm bounds on clients' updates. Based on this, we propose RoFL, a new secure FL system that extends secure aggregation with privacy-preserving input validation. 2/
The best and brightest minds in machine learning transcend borders and often follow different paths into research.
Proud to announce our C4AI Scholars program, which is part of our commitment to create more entry points into machine learning research.
https://t.co/nRtTWcBVOH
Very excited to present our work on Camel: Cryptographic Auditing for Collaborative Machine Learning at the @USENIXSecurity poster session together with my colleagues @KuechlerNicolas @lubux86@emanuel_opel@anwarhit Thanks everyone for coming by!
Today, at #OSDI21, @KuechlerNicolas and I will present Zeph, a system that enforces users' privacy preferences without trusting service providers with enforcement. Paper and artifact at: (https://t.co/iQEL25fCts). Work with @anwarhit, @AlexanderViand, @hossein_shafagh
Excited to share our upcoming @IEEESSP SoK paper on Fully Homomorphic Encryption Compilers, i.e. tools that help lower barriers of entry in FHE for non-expert developers. https://t.co/WfkKvJXOB8
Joint work with @pjattke@anwarhit 1/n