Privacy-Preserving Learning using Autoencoder-Based Structure

The need for privacy makes data centers not provide their datasets to inference centers. On the other hand, inference centers need more data to train learning algorithms and provide suitable and acceptable services. Therefore, the existence of a structure that can keep the data confidential while maintaining its usefulness for utility providers is of great importance. In this paper, by modifying the structure of the auto encoder, a method is presented that manages the trade-off between utility and privacy. Moreover, the performance of the proposed method has been evaluated by simulation.