Font Size: a A A

Research On Construction Process Of Privacy-Preserving Decision Tree

Posted on:2007-03-28Degree:MasterType:Thesis
Country:ChinaCandidate:H P LuFull Text:PDF
GTID:2178360182478498Subject:Pattern Recognition and Intelligent Systems
Abstract/Summary:PDF Full Text Request
In view of security, privacy preserving is becoming more and more highlighted in recent research of data mining. The paper briefly introduces the concept of privacy preserving data mining technology and studies the application of decision tree classifier in this particular field. A decision tree classifier is applied and Secure Two Party Computation is added, so that the need of privacy preserving is satisfied as well as the advantage of decision tree is retained.For privacy preserving, an improvement on the traditional decision tree algorithm is put forward. In the stage of choosing the best attribute in C4.5, we connect records in dataset with certain vectors, which helps incorporate Scalar Product Protocol into decision tree, and thus build a privacy-preserving decision tree classifier.Scalar Product Protocol is the keystone in building privacy-preserving decision tree. In view of the weakness of Semi-trusted Third Party Based Scalar Product Protocol, we bring forward a Preliminary Oblivious Transfer Based Scalar Product Protocol (Protocol 4.1), which is detached from a third party. With the help of Oblivious Transfer, our protocol can prevent one from knowing other's private data while doing scalar product calculation. We analyze the state transition for Protocol 4.1, and design an implementation for experiment. The experiment result shows our protocol is practical in both LAN and Internet. And comparing with the performance of simple Scalar Product Protocol without privacy preserving, the performance of Protocol 4.1 is acceptable.Based on Protocol 4.1, a high-performance Protocol 5.1 is put forwarded, with the name Oblivious Transfer Based Scalar Product Protocol. One vector is divided into m vectors, and thus privacy preserving effect is improved in geometric series, while the time overhead increases just a little. We analyze the state transition for Protocol 5.1, and design an implementation for experiment. The experiment result shows our protocol is practical in both LAN and Internet. And comparing with the performance of Protocol 4.1, the performance of Protocol 5.1 is satisfactory.
Keywords/Search Tags:Privacy, Scalar Product Protocol, Decision Tree, Oblivious Transfer
PDF Full Text Request
Related items