Supervisor:

Ph.D Helger Lipmaa (University of Tartu, Estonia)

Oponents:

Ph.D Carla Ràfols (Pompeu Fabra University, Barcelona, Spain)

Ph.D Ivan Visconti ( University of Salerno, Salerno, Italy)

Abstract:

In the current digital era, we can do increasingly astonishing activities remotely using only our electronic devices. Using mobile applications such as WhatsApp, we can contact someone with the guarantee, using an end-to-end encryption protocol, that only the recipient can know the conversation's contents. Most banking systems enable us to pay our bills and perform other financial transactions, and use the TLS protocol to guarantee that no one can read or modify the transaction data. Some countries provide an option to vote electronically in an election (e.g. Estonia) or referendum (e.g. Switzerland) with similar privacy guarantees to traditional paper voting. In all these activities, a cryptographic protocol is required to ensure users' privacy. In reality, some parties participating in a protocol might not act according to what was agreed in the protocol specification. Hence, for a real world protocol to be secure, we also need each party to prove that it behaves honestly, but without sacrificing privacy of its inputs. This can be done using a zero-knowledge argument: a proof by a polynomial-time prover that gives nothing else away besides its correctness. In many cases, we want a zero-knowledge argument to be non-interactive and transferable, so that it is computed only once, but can be verified by many verifiers at any future time. There are two main models that enable transferable non-interactive zero-knowledge (NIZK) arguments: the random oracle (RO) model and the common reference string (CRS) model. Protocols in the RO model are very efficient, but due to some of its limitations, we prefer working in the CRS model. In this work we provide three scenarios where NIZK arguments are relevant: verifiable computation, authorization, and electronic voting. In each scenario, we propose NIZK arguments in the CRS model that are more efficient than existing ones, and are comparable in efficiency to the best known NIZK arguments in the RO model.