Quantum-Resilient Distributed Learning on Neural-Networks
Event details
| Date | 30.07.2019 |
| Hour | 09:00 › 11:00 |
| Speaker | Sinem Sav |
| Location | |
| Category | Conferences - Seminars |
EDIC candidacy exam
Exam president: Prof. Carmela Troncoso
Thesis advisor: Prof. Jean-Pierre Hubaux
Co-examiner: Prof. Martin Jaggi
Abstract
The growing popularity of distributed machine learning services has increased the concerns about privacy and security as the data to be collected, shared, and trained is generally very sensitive. Organizations/individuals try to increase the size of the training set by collaboratively train their datasets. In this work, we propose a method for protecting the collaborative parties` data in training process. This method employs lattice-based cryptography and homomorphic operations on encrypted data.
Background papers
SecureML: A System for Scalable Privacy-Preserving Machine Learning, by P. Mohassel and Y. Zhang.
Helen: Maliciously Secure Coopetitive Learning for Linear Models, by Zheng, W., Popa, R. A., Gonzalez, J. E., Stoica, I.
Deep Learning with Differential Privacy, by Abadi, M., et al.
Exam president: Prof. Carmela Troncoso
Thesis advisor: Prof. Jean-Pierre Hubaux
Co-examiner: Prof. Martin Jaggi
Abstract
The growing popularity of distributed machine learning services has increased the concerns about privacy and security as the data to be collected, shared, and trained is generally very sensitive. Organizations/individuals try to increase the size of the training set by collaboratively train their datasets. In this work, we propose a method for protecting the collaborative parties` data in training process. This method employs lattice-based cryptography and homomorphic operations on encrypted data.
Background papers
SecureML: A System for Scalable Privacy-Preserving Machine Learning, by P. Mohassel and Y. Zhang.
Helen: Maliciously Secure Coopetitive Learning for Linear Models, by Zheng, W., Popa, R. A., Gonzalez, J. E., Stoica, I.
Deep Learning with Differential Privacy, by Abadi, M., et al.
Practical information
- General public
- Free