Privacy-Preserving Machine Learning for Generic Data

With the increasing use of personal data in recent years, privacy protection has become increasingly important. Differential privacy (DP) is a technique that protects privacy by adding noise. While DP manages privacy on a central server, Local Differential Privacy (LDP) can manage privacy locally, and LDP is preferable in terms of user privacy protection. In existing research, SUPM, a framework for applying LDP to machine learning, was proposed, which proposed WALDP, a privacy mechanism that treats all attributes uniformly regardless of data type, but it is a method specialized for continuous values. In this study, we propose a privacy-preserving machine learning method that can be applied to databases containing discrete values.