Secure computation is a technology that enables computation while keeping data encrypted. In the modern data analysis process, the data is transferred to, stored in, and computed by the server. During these three processes, a typical encryption protects the data only when it is transferred and stored. In contrast, secure computation protects the data during the whole processes: the data is kept secret even while it is computed. Therefore, secure computation not only prevents data leakage but also enables us to operate an analytical system that deals highly sensitive data, such as personal data and companies’ trade secrets, without data being disclosed. As a result, secure computation provides a novel analysis with integration of multiple data beyond the boundaries of companies and industries, as well as it provides highly secure data processing.
NTT’s secure computation uses secret sharing scheme that conforms to the ISO international standards as its encryption mechanism, and multiparty computation techniques based on the secret sharing scheme. Multiparty computation performs secure data processing while the data is encrypted by performing special cryptographic operations and exchanging encrypted data among multiple servers according to preliminarily defined procedures. All through these procedures, the data is processed in an encrypted fashion: each server only has a piece of the data called a share in the context of secret sharing scheme. Therefore, no server can obtain information about the data. Furthermore, we have achieved highly efficient data processing in secure computation, which is nearly as efficient as an ordinary (non-encrypted) data processing is.
San-Shi, the secure computation system developed by NTT, consists of three or four servers and a client that registers data via secret sharing scheme and instructs the servers to analyze the data. By using San-Shi, you can securely compute and obtain basic statistics and cross tabulation, etc. Furthermore, San-Shi provides a secure data integration by using a special secure table-join algorithm so you can perform a cross-section analysis of multiple data.
For further information about our secure computation, please see the downloadable material.