Browsing by Author "Ikonen, Antti-Jussi"
Now showing 1 - 1 of 1
- Results Per Page
- Sort Options
- Use of Zero-Knowledge Proofs in Federated Learning
Perustieteiden korkeakoulu | Bachelor's thesis(2024-05-07) Ikonen, Antti-JussiZero-knowledge proof is a cryptographic primitive that allows some statement to be proven true without disclosing any information about the content of the proof. Federated learning is a relatively new sub-field of machine learning, where the training of the global model is implemented decentralized, using clients' own devices. Since federated learning is a new sub-field of machine learning, there are several problems and challenges for which solutions are currently being sought. Some of these solutions utilize zero-knowledge proofs to some extent. The study investigates what different solutions utilizing zero-knowledge proofs have been proposed to address some of the challenges in federated learning. The study is a literature study, with the primary data consisting of scientific publications from recent years that present solutions utilizing zero-knowledge proofs to address some challenges in federated learning. In federated learning, clients, i.e., entities training local models, can prove to the server using zero-knowledge proofs that the local model has been trained correctly. The server can utilize zero-knowledge proofs to ensure clients that the aggregation of local models has been performed correctly and that no local models have been arbitrarily added, modified, or removed. Zero-knowledge proofs can also be used for client authentication and to prove ownership of models. As the field is very new, there is not yet available data on the utilization of zero-knowledge proofs in real-world applications. Further research on the topic can investigate whether zero-knowledge proofs are applicable in real-world scenarios or if their benefits remain theoretical.