Font Size: a A A

Research On The Key Issues Of Secure Outsourced Computation

Posted on:2016-06-24Degree:DoctorType:Dissertation
Country:ChinaCandidate:Q W LuFull Text:PDF
GTID:1228330467995016Subject:Information security
Abstract/Summary:PDF Full Text Request
With the advent of the era of big data, more and more data (e.g., social networking data, mobile positioning trajectory data, sensor sensing data, scientific observation data, etc.) are emerging into people’s daily life and work. The emergence of these massive, heterogeneous and inferior (inaccurate) data shows characteristics different from the past on volume, structure and quality. Meanwhile, with further pursue of the quality of daily life from people, computing tasks are becoming increasingly complex (e.g., more complex statistical, mining and scientific computing), whose computational overhead and resource consumption is constantly increasing. To meet this challenge by new data features and computing requirements, and obtain the results desired by the users both efficiently and accurately, due to the impossibility of traditional local computing mod-el (computing, storage resource-constrained, and the calculation is not economical) for these needs and the rapid development of cloud computing technology as the represen-tative of a third-party computing application, outsourced computation (outsourcing data to a third party to compute and obtain the return results) came into being.However, in the outsourced computing model supported by cloud computing tech-nology, the cloud may attack and corrupt the data privacy and the reliability of comput-ing results due to certain factors (e.g., hardware and software errors, extra commercial interests, etc.) Therefore, the study of security and privacy issues within outsourced computation is becoming increasingly crucial.In the thesis, we study several key issues in secure outsourced computation, and contribute the main work as follows:(1) We propose a privacy preserving trajectory data publishing scheme with per-sonalized privacy attributes. Facing the contradictory between data privacy and data availability, and observing the low data utility of most of existing publishing schemes with unified privacy protection requirement, we import the personalized privacy re-quirement for the data (such as record or trajectory) from different people, and study the personalized privacy-preserving data publishing mechanism. Specifically, duo to the trajectory urgent privacy protection need for mobile trajectory data, we study and propose the personalized trajectory data publishing scheme. We improve the utility of published data satisfying personalized privacy. The effectiveness and efficiency of our scheme is evaluated by experiment.(2) We propose an efficient verification scheme for uncertain frequent itemset min-ing based on power group construction and aggregate checking algorithm. Facing the contradictory between the reliability of computing results and limited control of com- puting, due to the increasingly complex data mining and computing task (e.g., frequent itsmest mining) and uncertain data (e.g., sensing data, statistic and probabilistic data), we study the verification mechanism for such complex computing task on uncertain da-ta. Since the important role of frequent itemset mining in data mining and statistic, we study the efficient verification scheme for uncertain frequent itemset mining. Specif-ically, we construct the power groups and build efficient checking algorithm for their verification. We prove the robustness and effectiveness of our scheme by theoretical proofs, and good efficiency by extensive evaluations.(3) We propose a secure collaborative computing scheme for multiple data own-ers based on multiplicative perturbation privacy protection. Facing the contradictory between data partition and secure collaborative computing among data owners, though there exists some research on secure oursourced computing, secure collaborative com-puting resulted from data partition has not attracted enough attention. Most of the schemes tackle such problem by cryptography and secure multiple party techniques, which leads to low efficiency and is not impractical. To solve this problem propose a secure collaborative computing scheme for multiple data owners based on multiplicative perturbation privacy protection. Specifically, we design the schemes in an incremental way for different adversary assumptions. We prove and evaluate the correctness and ef-fectiveness of out proposed schemes by the case study of several classical data mining task including KNN, K-means and SVM.
Keywords/Search Tags:Outsourced computation, Privacy-preserving, Data publishing, Personal-ization, Verification, Data partition
PDF Full Text Request
Related items