Privacy-Preserving Credit Scoring on Cloud

Credit scoring needs comprehensive data to achieve accurate assessment. However, these data often lie in the different places such as banks and financial institutions, internet firms, and almost all the data contain privacy information. Meanwhile, the acquisition of big data and privacy protection influence the rapid development of big data for credit scoring. And the introduction of big data for credit scoring proposes a lot of requirements for computing and storage capabilities. Cloud servers can provide powerful computing and storage services, but it also accompanies with higher privacy requirements. In this paper, we designed an additively homomorphic based secure multiparty computation scheme to collect and calculate credit data shared by different parties and at the same time preserve privacy in the cloud computing. We introduced two scenarios for credit scoring in this paper: one is to collect statistic information of relevant variables (such as a user’s overdue information in all banks) based on the existing credit model. The other is to collect a large amount of data for training to get credit evaluation model, but the efficiency of this scenario will be significantly lower due to the need of lots of multiplication operations. Finally, we analyzed the security and performance of our scheme, and proved that our scheme is safe and does not reveal the privacy of data in the cloud server.