Telecom Fraud Identification based on Vertical Federated Secure Boosting Tree Algorithm

With the continuous development of the Internet industry, the means of fraud have become more and more deceptive, and telecom fraud has gradually become a common means, telecom fraudsters are able to take advantages of new technological means to carry out criminal activities. Therefore, the model being able to accurately identify fraud phone numbers is demanded as it can fundamentally avoid the risk of fraud. However, the data of customers are always privacy sensitive. In order to prevent the privacy of each telecom operator’s customers from being leaked in the process of modeling whether an unknown phone number is a telecom fraud number using vertical federated learning. Secure Boosting Tree (SBT) algorithm was introduced to handle the phone number data, and the optimal model was obtained through a large number of comparative experiments. Compared with the results obtained by the traditional centralized learning method, the effectiveness of federated learning is verified.