Optimal interactive coding for insertions, deletions, and substitutions

Interactive coding, pioneered by Schulman (FOCS 92, STOC 93), is concerned with making communication protocols resilient to adversarial noise. The canonical model allows the adversary to alter a small constant fraction of symbols, chosen at the adversarys discretion, as they pass through the communication channel. Braverman, Gelles, Mao, and Ostrovsky (2015) proposed a far-reaching generalization of this model, whereby the adversary can additionally manipulate the channel by removing and inserting symbols. They showed how to faithfully simulate any protocol in this model with corruption rate up to 1/18, using a constant-size alphabet and a constant-factor overhead in communication. We give an optimal simulation of any protocol in this generalized model of substitutions, insertions, and deletions, tolerating a corruption rate up to 1/4 while keeping the alphabet to a constant size and the communication overhead to a constant factor. Our corruption tolerance matches an impossibility result for corruption rate 1/4 which holds even for substitutions alone (Braverman and Rao, STOC 11).