Enhancing Security Using Random Binary Weights in Privacy-Preserving Federated Learning

In this paper, we propose a novel method for enhancing security in privacy-preserving federated learning under the use of the vision transformer. In federated learning, learning is performed by collecting updated information without collecting raw data from each client. However, the problem is that raw data may be inferred from updated information. To address this issue, conventional data guessing countermeasures (security enhancement methods) have a trade-off relationship between privacy protection strength and learning efficiency, and generally degrade model performance. In this paper, we propose a novel method of federated learning that does not degrade model performance and is robust against data guessing attacks on updated information. In the proposed method, each client independently prepares a sequence of binary (0 or 1) random numbers, multiplies it by the update information, and sends it to the server for model learning. In experiments, the effectiveness of the proposed method is confirmed in terms of model performance and resistance to the APRIL (Attention PRIvacy Leakage) restoration attack.