Abstract
Secure computation enables mutually distrusting parties to jointly evaluate a function on their private inputs without revealing anything but the function’s output. Generic secure computation protocols in the semi-honest model have been studied extensively and several best practices have evolved.
In this work, we design and implement a mixed-protocol framework, called ABY, that efficiently combines secure computation schemes based on Arithmetic sharing, Boolean sharing, and Yao's garbled circuits and that makes available best practice solutions in secure two-party computation. Our framework allows to pre-compute almost all cryptographic operations and provides novel, highly efficient conversions between secure computation schemes based on pre-computed oblivious transfer extensions. ABY supports several standard operations and we perform benchmarks on a local network and in a public intercontinental cloud. From our benchmarks we deduce new insights on the efficient design of secure computation protocols, most prominently that oblivious transfer-based multiplications are much more efficient than multiplications based on homomorphic encryption. We use ABY to construct mixed-protocols for three example applications - private set intersection, biometric matching, and modular exponentiation - and show that they are more efficient than using a single protocol.
Joint work with Daniel Demmler and Michael Zohner published at NDSS 2015.