Data Anonymnization for Privacy Protection Using Deep Learning Techniques

Abstract

The majority of real data are unbalanced, which makes them challenging to classify. Unbalanced data has been the subject of some research. However, deep learning has not taken into account how to manage such data. In this paper, we suggest deep learning-based data mining with imbalanced data for privacy protection. We discuss and analyse the aspects of current privacy-protection data mining as well as a deep learning based anonymizing tool. Analyzing data anonymously does not affect deep learning's accuracy, according to experiments employing various anonymization technologies. Undersampling and upper-sampling are both employed for imbalanced data; however, it is unknown whether they have a tendency to lower data accuracy because of anonymization.