Design And Implementation Of Fast Proof Generation System For Verifying Cloud Search

Nowadays cloud computing is popular in business and daily usage. More and more documents are kept in cloud storage for large storage capicity, high availability and accessibility. To providing better service, searching documents on cloud is provided by cloud providers. How to proof the correctness of the searching result is an urgent problem coming with this feature. Previous work has proposed many useful theories on verifiable computing, such as probabilistically checkable proofs, authHashtable, but an efficient and scalable system is rare.In our research, first, we design and implement a fast proof generation system for verifying cloud search based on RSA accumulators, membership witnesses and nonmembership witnesses. We split our system into modules and every module is a process that makes out system have high expansibility and maintainability.Then, in order to reduce the correctness proof generation time of large document set, we design and implement the tree-structure proof which splits the large set into multiple small sets to speed up the computing. We also present a method of using tree-structure to calculate integrity proof. We implement a lightweight parallel computing framework which can easily use different number of computers to calculate the tree-structure proof. Finally, we design a probabilistically checkable method based on sampling. This methed lets user can make tradeoff between performace and assurance level.Experimental results show that by using the tree-structure proof the correctness proof generation speed has an average speedup of 2.5. For the overall performance, our system has a speedup of 2 compared to the straightforward way. Our system can handle dataset with a size of 100 GB in our experimental environment.