Jacopo Giglioli
Privacy-Preserving Federated Learning with Homomorphic Encryption on GPU.
Rel. Valentino Peluso, Erich Malan, Erkay Savas. Politecnico di Torino, Corso di laurea magistrale in Cybersecurity, 2026
|
Preview |
PDF (Tesi_di_laurea)
- Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (5MB) | Preview |
Abstract
Federated Learning is a distributed machine learning paradigm that enables multiple clients to jointly train a global model while keeping each client's dataset local and private. In each training round, clients download the current global model, fine-tune it locally using their data and send the resulting updates back to the server, which aggregates them to produce a new version of the global model. Although not transmitting clients' raw data already represents a significant improvement for privacy, some threats still exist, as an attacker can still infer sensitive information about clients' datasets from the shared updates. Homomorphic Encryption, which enables computation on encrypted data without requiring prior decryption, offers a robust solution to this concern.
The communication channel between clients and server is encrypted, preventing an attacker from reading the updates shared by the nodes
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Corso di laurea
Classe di laurea
Ente in cotutela
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
