Information-Theoretic Security of Cryptographic Channels

Abstract

We discuss the setting of information-theoretically secure channel protocols where confidentiality of transmitted data should hold against unbounded adversaries. We argue that there are two possible scenarios: One is that the adversary is currently bounded, but stores today's communication and tries to break confidentiality later when obtaining more computational power or time. We call channel protocols protecting against such attacks future-secure. The other scenario is that the adversary already has extremely strong computational powers and may try to use that power to break current executions. We call channels withstanding such stronger attacks unconditionally-secure. We discuss how to instantiate both future-secure and unconditionally-secure channels. To this end we first establish according confidentiality and integrity notions, then prove the well-known composition theorem to also hold in the information-theoretic setting: Chosen-plaintext security of the channel protocol, together with ciphertext integrity, implies the stronger chosen-ciphertext notion. We discuss how to build future-secure channel protocols by combining computational message authentication schemes like HMAC with one-time pad encryption. Chosen-ciphertext security follows easily from the generalized composition theorem. We also show that using one-time pad encryption with the unconditionally-secure Carter-Wegman MACs we obtain an unconditionally-secure channel protocol.