Prevention of GAN-Based Privacy Inferring Attacks Towards Federated Learning

With the increasing amount of data, data privacy has drawn great concern in machine learning among the public. Federated Learning, which is a new kind of distributed learning framework, enables data providers to train models locally to protect privacy. It solves the problem of privacy leakage of data by enabling multiple parties, each with their training dataset, to share the model instead of exchanging private data with the server side. However, there are still threats of data privacy leakage in federated learning. In this work, we are motivated to prevent GAN-based privacy inferring attacks in federated learning. For the GAN-based privacy inferring attacks, inspired by the idea of gradient compression, we propose a defense method called Federated Learning Parameter Compression (FLPC) which can reduce the sharing of information for privacy protection. It prevents attackers from recovering the privacy information of victims while maintaining the accuracy of the global model. Comprehensive experimental results demonstrated that our method is effective in the prevention of GAN-based privacy inferring attacks.