Secure multi-party protocol using modern cryptographic technique and tamper resistant hardware

Abstract

Secure Multi-party Computation (SMC) is one significant research area in information security. In SMC, multiple parties jointly work on some function and no parties take the risk of revealing their private data. Since A.C. Yao first proposed this problem in 1982, there have been a lot of researchers working on different versions of SMC. In this thesis, we address three different researches in this setting, including the Privacy-Preserving Cooperative Scientific Computation, Privacy Preserving Data Mining (PPDM), and PPDM in cloud environment.

In Privacy-Preserving Cooperative Scientific Computation, we propose a solution to the Privacy Preserving Weighted Average Problem (PPWAP) under the hybrid security model, which guarantees the malicious parties will not get the correct final result if they behalf maliciously. Later, the extended version of our scheme is shown as a highly efficient and secure PPWAP solution under the malicious model, a stronger security model requiring more resource.

Privacy reserving data mining is one important branch of SMC, where all participants want to get the same and correct mining result from collaborated data mining without any threat of disclosing their private data. In another word, each party refuses to review its individual private database while carrying out collaborated data mining. We propose a PPDM solution of building up a decision tree from a hybrid distributed database, which is a quite common situation in real life but has not been solved before. Previous research works only focus on horizontally or vertically distributed database. With the great development of cloud computing, it provides a much more flexible and efficient platform for Internet service providers and users. However, the privacy issues of cloud service has become the bottleneck of its further development, and this problem also draw a lot of researchers' attention in recent decade. In this thesis, we propose the first solution to cloud-based PPDM. The cloud server carries out data mining on encrypted databases, and our solution can guarantee the privacy of each client. This scheme can protect client from malicious users. With aid of a hardware box, our design can also protect clients from untrusted cloud server. Another novel feature of this solution is that it works even when the databases from different parties share overlapped parts. Furthermore, with the help of homomorphic encryption and black box, our scheme can carry out the calculation on the overlapped data. This kind of problem has never been resolved by previous works as far as we know.