Perfect Secrecy #
Characterisation theorems for perfect secrecy following [KL20], Chapter 2: the equivalence with message-ciphertext independence, the ciphertext indistinguishability characterization, and Shannon's key-space bound.
Main results #
Cslib.Crypto.Protocols.PerfectSecrecy.EncScheme.perfectlySecret_iff_indep: perfect secrecy is exactly message-ciphertext independenceCslib.Crypto.Protocols.PerfectSecrecy.EncScheme.perfectlySecret_iff_ciphertextIndist: ciphertext indistinguishability characterization ([KL20], Lemma 2.5)Cslib.Crypto.Protocols.PerfectSecrecy.EncScheme.perfectlySecret_keySpace_ge: Shannon's theorem,|K| ≥ |M|([KL20], Theorem 2.12)
References #
The joint distribution at (m, c) equals msgDist m * ciphertextDist m c.
Summing the joint distribution over messages gives the marginal ciphertext distribution.
Perfect secrecy is equivalent to message-ciphertext independence.
The two formulations are related by multiplying/dividing by marginal(c).
A scheme is perfectly secret iff the ciphertext distribution is independent of the plaintext ([KL20], Lemma 2.5).
Ciphertext indistinguishability implies message-ciphertext independence.
Perfect secrecy requires |K| ≥ |M| — Shannon's theorem
([KL20], Theorem 2.12).