Abstract
In this seminar, we will provide an overview of totally homomorphic encryption (FHE), focusing on schemes based on the Learning With Errors (LWE) problem . In particular, we will study the main operations of the TFHE scheme and its bootstrapping. We then compare the two main families of LWE-based FHE schemes: the BGV family, including CKKS, and the GSW family, including TFHE. We'll analyze the advantages and disadvantages of each, and show how both approaches can be implemented to homomorphically evaluate an active artificial neuron. A short demonstration will be given using the DESILO FHE library. Finally, we will conclude with some thoughts on the practical use of FHE and its interaction with other PrivacyEnhancing Technologies.
