Computational Wiretap Coding from Indistinguishability Obfuscation

Explore computational wiretap coding schemes that enable secure communication over noisy channels using indistinguishability obfuscation in this 32-minute conference talk. Learn how Alice can reliably transmit messages to Bob through one noisy channel while simultaneously hiding the same message from an adversary Eve who receives the encoding through a different noisy channel. Discover the fundamental nondegradation condition that requires Bob's channel not to be a degradation of Eve's channel, meaning Eve cannot simulate Bob's view. Examine how this condition, while insufficient in information-theoretic settings, becomes sufficient in computational settings under cryptographic assumptions. Understand the breakthrough result that settles the feasibility question for all discrete memoryless channels with binary input alphabets and arbitrary finite output alphabets, based on standard sub-exponential hardness assumptions including those implying indistinguishability obfuscation and injective pseudorandom generators. Analyze the specific application to binary symmetric channels and binary erasure channels, where secure communication is feasible when the erasure probability exceeds twice the crossover probability. Gain insights into the new polytope characterization of channel degradation for binary-input channel pairs that contributes to the information-theoretic foundations of this work.