A party-adaptive variant of LowGear

Abstract Over the last decade, secure multi-party computation (MPC) has made a great stride from a major theoretical area to the multi-functional tool for building privacy protection applications. At EUROCRYPT’18, Keller et al. presented MPC protocol LowGear based on semi-homomorphic (addition-only) encryption. Due to its superior performance and high availability, LowGear has drew a lot of attention from industry. But in practical application environment, there are still requirements that LowGear does not satisfy. A major drawback is the inability to handle changes in the set of parties during online computing phase. In this paper, we present a party-adaptive variant to LowGear, with the goal to continue secure collaborative computing tasks when some new parties join in, some old parties drop out, or new parties replace old ones. This variant is based on tweaks to preprocessing phase to adapt existing preprocessed data to the computation over a new set of parties without regenerating new data by rerunning the entire preprocessing phase. It can effectively avoid the waste of data and time resources. Meanwhile the variant does not reduce security and efficiency of LowGear.