Using fully linear zero-knowledge proof systems for practical “semi-honest to malicious” compilers with sublinear additive communication overhead.Īdria Gascon, Peter Kairouz, Kallista (Kaylee) Bonawitz: Privacy in Federated Learning at Scale Using pseudorandom correlation generators for “silent” secure generation of useful sources of correlated randomness, including random oblivious transfers and (authenticated) multiplication triples This applies in particular to ReLU and other popular activation functions Using function secret sharing for fast secure computation of simple nonlinear functions in an offline-online setting. In this talk I will survey recent techniques that were developed in the context of general-purpose secure computation but can be particularly appealing for ML-related applications: The explosive growth of interest in privacy-preserving ML motivates new techniques for efficient secure computation.
![crypto bar ilan crypto bar ilan](https://crypto.ku.edu.tr/wp-content/uploads/2018/11/eyl%C3%BCl3.jpg)
Nicolas Papernot (University of Toronto)Ībstracts – Invited Speakers Yuval Ishai: New Techniques for Efficient Secure Computation.Videos All videos of the workshop are available at the IACR You Tube Channel here Location Zoom. Please read the participation guidelines.
CRYPTO BAR ILAN REGISTRATION
To register to the workshop, please register to CRYPTO 2021, and mark in the registration form the PPML workshop. The workshop is an affiliated event of CRYPTO 2021.
![crypto bar ilan crypto bar ilan](https://www.unboundsecurity.com/wp-content/uploads/2021/07/Supply-Chain-Attacks_800-768x399.png)
The workshop will consist of few invited talks, together with contributed talks. The scope includes privacy preserving techniques for training, inference, and disclosure. The workshop aims to strengthen collaborations among the machine learning and cryptography communities. Can we train machine learning algorithms on confidential data without ever being exposed to it? Can my model classify your sample without ever seeing it? Machine learning algorithms perform better when being exposed to more and more data, but such data is not always accessible due to privacy constraints. Applications of machine learning involve almost every aspect of our lives, from health care and DNA sequence classification, to financial markets, computer networks and many more. Systems based on machine learning algorithms approach and sometimes even exceed the abilities of human experts. Affiliated Event: The 3rd Privacy-Preserving Machine Learning Workshop 2021 AboutĪrtificial intelligence is progressing rapidly.