1 January 2025 in Nonlinear Theory and Its Applications, IEICE
This paper proposes an autonomous and secure distributed Neural Gas (NG) method using decomposed data on multiple servers uniformly distributed in a ring shape and connected only to neighboring servers. The advantages of the proposed method are that learning can be performed with decomposed data and parameters, thus ensuring the confidentiality of the data and parameters; the distributed processing system is easy to connect; and servers can be easily added or removed because of the uniform structure where all servers perform the same operations. There are two types of learning targeted by the secure distributed processing: supervised and unsupervised learning. In the previous paper, we proposed the Back Propatation (BP) method as an example of the former. Here, we propose the NG and k-means methods as examples of the latter. The advantage of unsupervised learning is that it can discover trends and segments found in given data without labels (correct answers) for machine learning. Since there is no need to obtain correct answer data in advance, this learning method can be used for a wider range of tasks. The effectiveness of the proposed NG and k-means methods of secure distributed processing is demonstrated by comparing it with conventional methods through numerical simulations of clustering.
1 January 2023 in Nonlinear Theory and Its Applications, IEICE
Much research on secure and safe AI analysis methods for users has been conducted. They are 1) secure multiparty computation (SMC), 2) quasi-homomorphic encryption, 3) federated learning (FL), and so on. It is known that both utility and confidentiality are essential for machine learning using confidential data. However, there is a trade-off between them. In the previous paper, we proposed a learning method for secure distributed processing using decomposition data and parameters. The characteristic feature of this method is to achieve high confidentiality through distributed processing with decomposed data and parameters. On the other hand, the realization of machine learning through the distribution and integration of decomposition data leads to an increase in computational complexity and a deterioration of computational accuracy. In particular, while increasing the number of servers improves confidentiality, this problem becomes more pronounced. In this paper, we propose scalability improvement of simplified, secure distributed processing methods of BP and NG methods to suppress the computational complexity associated with an increase in the number of servers and demonstrate its effectiveness through a numerical simulation.