Online consecutive secure multi-party computation algorithm for preserving privacy

• Main Author: Kenari, Abdolreza Rasouli (Author)
• Format: Thesis
• Published: 2011-11.
• Subjects: QA75 Electronic computers. Computer science; Thesis
• Online Access: Get fulltext

 Every day large volume of information produces and stores among multi parties systems. Although these data are produced by companies unrelated to each other and are stored in various parties, but when they are gathered together much valuable information and patterns reveals. Data mining over distributed data discovers this costly knowledge. However, ownership of data by different companies and maintain confidentiality of data is the main challenge in this research. Secure Multi-party Computation is a set of methods that perform mathematic computation over the multi-party distributed data with ensuring the privacy preserving of the confidential data. Most of these methods use a shared secret key to ensure the privacy of each party. All parties should be present to share the secret keys, but unfortunately, in many applications, the parties are not joining the process at the start time. The aim of this research is to design a new online consecutive secure multiparty computation algorithm. The main problems addressed in this research are the online secret sharing and the consecutively two-party computation. The infinite convergent product sequences are employed to overcome the dependency problem between the shared secret key and users' public keys, which make the algorithm runs offline. The designed online secret sharer allows the parties to join the system during process life. The second problem is cleared by adding a two-party computation randomizer to the system. The two-party randomizer ensures the privacy of the online consecutive computation. The designed algorithm is tested and the result proves the security and applicability of the algorithm. Moreover, a distributed online frequent itemset mining is developed using the proposed algorithm and the result demonstrates the performance, efficiency and practicability of the multi-party computation algorithm. The result shows that the algorithm lasts only 0.5 second for thousand of the parties in offline mode and 27 minutes in the case of online mode with millions of participants.