Efficient Secure Computation and Randomness

The question of how to construct optimally efficient secure protocols is a central question in cryptography and in the computer security world at large. We focus in this work on several facets of this question, with a particular view towards the role of randomness in secure computation.;The use of random inputs is ubiquitous in cryptographic primitives. However, the ability to consistently draw from large random sources may be a barrier in practice. We examine the existence of optimally strong pseudorandom sources with particularly efficient implementations; namely, we construct exponentially hard pseudorandom generators that can be computed by circuits that have size linear in the generator output.;Conversely, we also examine efficient protocols that rely only minimally on their random inputs. In the resettable security model, parties may be forced to use the same random input across polynomially many interactions with other parties. We examine this security model in the case of zero knowledge proofs, which are a primitive frequently required in secure computation protocols when one party must prove that they have executed the protocol correctly to another party without revealing secret inputs and compromising security.;Finally, we examine the secure and efficient implementation of a specific functionality, including its various required cryptographic primitives (such as zero knowledge arguments of knowledge). In particular, we construct a protocol that securely realizes pattern matching, including single character wildcards and substring matching.