Cliptography: Post-Snowden Cryptography

Dr Qiang Tang
Computer Science Department, NJIT


Abstract

Despite the laudatory history of development of modern cryptography, an implicit assumption that the implementations are trusted has been made in (essentially) all security definitions. While the dangers of entertaining adversarial implementation of cryptographic primitives seem obvious, the ramifications of such attacks are surprisingly dire: it turns out that—in wide generality—adversarial implementations of cryptographic (randomized) algorithms may leak private information while producing output that is statistically indistinguishable from that of a faithful implementation. I will first briefly describe the intuition about the above subliminal channel attacks. Inspired by some folklore practical wisdom, I will also introduce some simple but rigorous immunizing strategies on subverted randomized algorithms.