Privacy-Preserving Deterministic Automata Evaluation with Encrypted Data Blocks

Secure computation (i.e., performing computation while keeping inputs private) is a fundamental problem in cryptography. In this paper, we present an efficient and secure 2-party computation protocol for deterministic automata evaluation, a problem of large practical relevance. Our result is secure under standard assumptions and bypasses roadblocks in previous general solutions, like Yao’s garbled circuits and Gentry’s lattice-based fully homomorphic encryption, by performing secure computations over data blocks (instead of bits) and using typical-size (instead of impractically large) cryptographic keys. An important efficiency property achieved is that the number of both asymmetric and symmetric cryptographic operations in the protocol is sublinear in the size of the circuit representing the computed function (specifically, improving linear-complexity protocols by a multiplicative factor equal to a block size). All previous protocols for deterministic automata evaluation required a linear number of asymmetric cryptographic operations. Moreover, we use quantitative comparison techniques to show that in typical parameter settings, our protocols’ latency is at least 1 to 2 orders of magnitude smaller than the protocol obtained by a direct application of both state-of-the-art general-purpose secure 2-party computation protocols. Even though not as general as in these two general-purpose techniques, our result is applicable to the class of all constant-space computations.