Building oblivious transfer on channel delays

In the information-theoretic setting, where adversaries have unlimited computational power, the fundamental cryptographic primitive Oblivious Transfer (OT) cannot be securely achieved if the parties are communicating over a clear channel. To preserve secrecy and security, the players have to rely on noise in the communication. Noisy channels are therefore a useful tool to model noise behavior and build protocols implementing OT. This paper explores a source of errors that is inherently present in practically any transmission medium, but has been scarcely studied in this context: delays in the communication. In order to have a model for the delays that is both general and comparable to the channels usually used for OT - such as the Binary Symmetric Channel (BSC) - we introduce a new noisy channel, the Binary Discrete-time Delaying Channel (BDDC). We show that such a channel realistically reproduces real-life communication scenarios where delays are hard to predict and we propose a protocol for achieving oblivious transfer over the BDDC. We analyze the security of our construction in the semi-honest setting, showing that our realization of OT substantially decreases the protocol sensitivity to the user's knowledge of the channel compared to solutions relying on other channel properties, and is very efficient for wide ranges of delay probabilities. The flexibility and generality of the model opens the way for future implementation in media where delays are a fundamental characteristic.