Privacy-Preserving Data Mining in Electronic Surveys

Electronic surveys are an important resource in data mining. However, how to protect respondents' data privacy during the survey is a challenge to the security and privacy community. In this paper, we develop a scheme to solve the problem of privacy-preserving data mining in electronic surveys. We propose a randomized response technique to collect the data from the respondents. We then demonstrate how to perform data mining computations on randomized data. Specifically, we apply our scheme to build a Naive Bayesian classifier from randomized data. Our experimental results indicate that accuracy of classification in our scheme, when private data is protected by randomization, is close to the accuracy of a classifier build from the same data with the total disclosure of private information. Finally, we develop a measure to quantify privacy achieved by our proposed scheme.

• Type: Conference
• Year: 2004
• Where: ICEB
• Authors: Justin Z. Zhan, Stan Matwin
• Keywords: Business, Data Mining, Electronic Surveys, ICEB 2004, Privacy-preserving Data Mining