The one-time pad revisited

The one-time pad, the mother of all encryption schemes, is well known to be information-theoretically secure, in contrast to most encryption schemes used in practice, which are at most computationally secure. In this paper, we focus on another, completely different aspect in which the one-time pad is superior to normal encryption, and which surfaces only when the receiver (not only the eavesdropper) is considered potentially dishonest, as can be the case in a larger protocol context in which encryption is used as a sub-protocol. For example, such a dishonest receiver (who is, say, coerced by the eavesdropper) can in normal encryption verifiably leak the message to the eavesdropper by revealing the secret key. While this leakage feature can provably not be avoided completely, it is more limited if the one-time pad is used. We use the constructive cryptography framework to make these statements precise.