Secure Multiparty Computation of Chi-Square Test Statistics and Contingency Coefficients

摘要

Generally, in order to perform data mining, all the original data should be provided to the third party first. However, in case of privacy-preserving data mining, the data provider may not want to disclose sensitive data directly to the third party. Therefore, it is very important to compute chi-square test statistics and contingency coefficients, which are statistically very useful, while not disclosing original sensitive data. In this paper, we propose a novel solution to securely compute chi-square test statistics and contingency coefficients by using secure scalar products and proposing secure bitmap string operations. We prove the correctness and secureness of the proposed solution by presenting a formal theorem. Also, we empirically show the superiority and practicality of the proposed method.