Using Secure Enclaves For Efficient Multi-party Computation

Secure multi-party computation(MPC),which enables distrust parties to compute a joint function on their private inputs,is a hot topic in the field of cryptography.Based on different security assumptions of participants,MPC is classified as the protocol with semi-honest security and the one with malicious security.The semi-honest assump-tion is hard to hold in most real world applications.In practice,malicious participants can actively attack the protocol to compromise the privacy of honest ones.However,MPC protocols will incur a significant overhead to defend against malicious partici-pants,and thus they are impractical.Furthermore,even MPC in the semi-honest model is unable to handle a data-intensive secure computation.To address these problems,we introduce SGX(Software Guard Extensions)secure enclave into MPC to reduce the overhead of traditional MPC protocols.Additionally,participants will run different SGXs participate in MPC,which we call heterogeneous SGXs.The mutual attestation between heterogeneous SGXs is not well supported by the existing remote attestation mechanism.The main contributions of this thesis are as follows:First,we propose a mutual attestation protocol for heterogeneous SGXs utilized in MPC.Heterogeneous SGXs that running different programs can cooperate with each other to finish secure computing tasks.This will save enclave memory and improve the performance of MPC.However,without the assistance of a third party,it is im-possible to do mutual attestation between heterogeneous SGXs by using the existing SGX mechanism.In order to address this problem,based on the characteristics of mu-tual attestation MRENCLAVE generation,we propose a mutual attestation protocol for heterogeneous SGXs.In this way,the security risk introduced by the third party in the traditional scheme is avoided.Secondly,based on SGX we devise an efficient two-party secure computation pro-tocol under the semi-honest assumption.Observed that participants have different pri-vacy concerns regarding different data.Participants can process different secret level data with diffrent approaches.In addition,due to side-channel threats of SGX,the se-curity level of SGX and two-party secure computation protocol is different,but the overhead of SGX is much lower than that of the two-party secure computation proto-col.To improve the overall performance,Our efficient two-party secure computation protocol works as follows:the secret data is processed by SGX and the top secret data is computed by a two-party secure computation protocol.Finally,utilizing SGX we design an efficient multi-party secure computation pro-tocol under the malicious assumption.First of all,we only assume a minimal trust on SGX which guarantees integrity of computation.Based on this assumption,starting with a two-party secure computation with malicious participants,By regarding SGX as a computational aid,we design an efficient two-party secure computation protocol.Finally,on using SGX to bound the malicious behavior of participants,we extend the two-party secure computation to a general multi-party secure computation.