no code implementations • 15 Nov 2023 • Itamar Zimerman, Moran Baruch, Nir Drucker, Gilad Ezov, Omri Soceanu, Lior Wolf
This innovation enables us to perform secure inference on LMs with WikiText-103.
no code implementations • 26 Apr 2023 • Moran Baruch, Nir Drucker, Gilad Ezov, Yoav Goldberg, Eyal Kushnir, Jenny Lerner, Omri Soceanu, Itamar Zimerman
Training large-scale CNNs that during inference can be run under Homomorphic Encryption (HE) is challenging due to the need to use only polynomial operations.
no code implementations • 3 Nov 2020 • Ehud Aharoni, Allon Adir, Moran Baruch, Nir Drucker, Gilad Ezov, Ariel Farkash, Lev Greenberg, Ramy Masalha, Guy Moshkowich, Dov Murik, Hayim Shaul, Omri Soceanu
We present a simple and intuitive framework that abstracts the packing decision for the user.
1 code implementation • 6 Aug 2020 • Abigail Goldsteen, Gilad Ezov, Ron Shmelkin, Micha Moffie, Ariel Farkash
The EU General Data Protection Regulation (GDPR) mandates the principle of data minimization, which requires that only data necessary to fulfill a certain purpose be collected.
1 code implementation • 26 Jul 2020 • Abigail Goldsteen, Gilad Ezov, Ron Shmelkin, Micha Moffie, Ariel Farkash
Anonymized data, however, is exempt from the obligations set out in these regulations.
no code implementations • 29 Jun 2020 • Abigail Goldsteen, Gilad Ezov, Ariel Farkash
These attacks are able to reveal the values of certain sensitive features of individuals who participated in training the model.