Miloslav Dušek

Oblivious transfer (OT) is a cryptographic primitive which is universal for multiparty computation. Unfortunately, perfect information-theoretically secure (ITS) quantum oblivious transfer is impossible. Imperfect information-theoretically secure quantum oblivious transfer is possible, but the smallest possible cheating probabilities are not known. We present an imperfect information-theoretically secure quantum oblivious transfer protocol with no restrictions on dishonest parties, and its experimental implementation. The cheating probabilities are 0.75 and 0.729 for sender and receiver respectively, which is lower than in existing protocols. Using a photonic test-bed, we have implemented the protocol with honest parties, as well as optimal cheating strategies.

Oblivious transfer (OT) is an important cryptographic primitive for transmitting information between two non-trusting parties and can be used as basic building block to implement any two-party computation. One variant of OT is XOR oblivious transfer (XOT), where the sender Alice has two bits and sends them to the receiver Bob. Bob will obtain either the first bit, the second bit, or their XOR. In an honest run of the protocol, Bob should not learn anything more than this, and Alice should not be able to tell what Bob has learned. Unfortunately, perfect quantum OT is impossible with information-theoretic security, so we focus on obtaining the smallest possible cheating probabilities for dishonest parties, when there are no restrictions imposed on them. We present a non-interactive quantum XOT protocol with classical post-processing, where the cheating probabilities are 1/2 for Alice and 3/4 for Bob. Reversing this protocol, so that Bob becomes the sender of a quantum state and Alice the receiver who measures it, while still implementing oblivious transfer from Alice to Bob, we show that the cheating probabilities for both parties stay the same as for the unreversed protocol. The reversed protocol is even easier to implement. The quantum XOT protocol outperforms classical XOT protocols. Lastly, we are in the process of implementing both the unreversed and the reversed protocol experimentally.