Title:	Security of Fully Homomorphic Encryption for privacy-preserving machine learning
DNr:	Berzelius-2026-179
Project Type:	LiU Berzelius
Principal Investigator:	Thomas Johansson <thomas.johansson@eit.lth.se>
Affiliation:	Lunds universitet
Duration:	2026-06-26 – 2027-01-01
Classification:	10211
Keywords:

Abstract

In order to assess the security level of different parameter choices in fully homomorphic encryption (FHE) as well as different FHE algorithms, we want to test the performance of at least two new solving algorithms. One is a meet-in-the-middle approach for solving RLWE with low weight secret, the second one is a noise-flooding attack on a recently proposed multi-key FHE scheme. Both tasks require large computational resources.