An introduction to Fully Homomorphic Encryption and its application to Machine Learning
Tesi esterna in azienda
Parole chiave FULLY HOMOMORPHIC ENCRYPTION
Riferimenti DANILO BAZZANELLA
Riferimenti esterni Guglielmo Morgari (Telsy)
Gruppi di ricerca Crittografia e teoria dei numeri
Tipo tesi TESI DI RICERCA IN AZIENDA
Descrizione Scenario: With the term Fully Homomorphic Encryption (FHE) we denote a set of cryptosystems that allow us to perform computations directly on encrypted data, hence guaranteeing the confidentiality of data-in-use. After the first FHE scheme was described in 2009 by Craig Gentry, we have assisted to the proposal of numerous Homomorphic Encryption schemes (TFHE, CKKS, BGV) that drastically improved performance. These schemes, together with software libraries developed by academic researchers and tech companies like Microsoft and IBM, allow users to use FHE schemes in real application scenarios like Machine Learning, medical research and e-voting.
Thesis proposal: After an introduction in which we will give a precise definition of Fully Homomorphic Encryption, we are going to mainly focus on the possible applications of FHE schemes to real world scenarios and in particular to Machine Learning. After an analysis of the available FHE libraries, we will develop a Proof-of-Concept implementation of a Machine Learning algorithm that, by making use of Fully Homomorphic Encryption, can be applied directly to ciphertexts and is thus able to perform all the needed computations without requiring access to the decrypted data.
Scadenza validita proposta 20/02/2024 PROPONI LA TUA CANDIDATURA