摘要：To enable data analytics that provides valuable insights, data that are distributed acrossseveral organisations increasingly need to be shared before they can be analysed. However, sharingdata from different sources can raise privacy and confidentiality concerns. Organisations are oftenunwilling or not allowed to share their sensitive data, such as personal details or health or financialdata, with other parties because this potentially violates the privacy of individuals. Secure multi-partycomputation (SMC) has been introduced as a solution to overcome the problem of performing computationson sensitive data across organisations. SMC allows parties to jointly compute a functionover their inputs while preserving the privacy of these inputs. Secure summation protocols are an importantbuilding block in many SMC applications that can be used under two different SMC models(i.e. with and without the involvement of a third party to conduct the computations). A secure summationprotocol is used to compute the summation of private inputs held by different parties. In thispaper we study existing secure summation protocols that can be used under different SMC modelsand then propose three advanced secure summation protocols that use homomorphic encryption. Wethen consider different scenarios of how parties might collude with each other in secure summationprotocols, and the potential collusion risks that occur with these protocols. No such investigation ofpossible collusion scenarios for secure summation protocols has so far been presented. We analyseeach secure summation protocol under different collusion scenarios and evaluate the efficiency ofeach protocol with different numbers of parties and different input data sizes. Our evaluation showsthat our proposed protocols provide improved privacy against collusion risks and they can calculatea sum more efficiently compared to existing secure summation protocols..
