Training method and device of longitudinal federal learning model and computer equipment

Abstract

The invention discloses a training method and device for a longitudinal federated learning model, and relates to the field of big data and the technical field of computers such as deep learning. According to the specific implementation scheme, the training method is applied to first participant equipment, and the first participant equipment has label data. The longitudinal federated learning model comprises a first bottom layer sub-model in the first participant device, an interaction layer sub-model, a top layer sub-model based on a Lipschitz neural network and a second bottom layer sub-model in the second participant device. And obtaining first bottom layer output data of the first participant device and second bottom layer output data sent by the second participant device, inputting the first bottom layer output data and the second bottom layer output data to the interaction layer sub-model to obtain interaction layer output data, obtaining top layer output data according to the interaction layer output data and the top layer sub-model, and sending the top layer output data to the interaction layer sub-model. And training a longitudinal federal learning model according to the top output data and the label data. According to the invention, the anti-attack capability of the federal learning system can be improved.

Description

technical field [0001] The present application relates to the field of computer technology, especially to the field of big data and deep learning, and in particular to a training method, device, computer equipment and storage medium for a longitudinal federated learning model. Background technique [0002] Federated learning is suitable for large-scale distributed deep learning model training, while providing privacy protection, and building a joint model for local data sets. In the vertical federated learning scenario, the dataset is split vertically and owned by different parties, that is, each party owns a disjoint subset of attributes, and the goal is to transfer data from one party to another without transferring any data. Collaborative learning machine learning models under the premise of the party. [0003] In related technologies, most of the security and privacy protection schemes in the vertical federated learning scenario focus on solving the privacy leakage prob...

AI Technical Summary

Problems solved by technology

However, in the actual application scenario, the participating nodes are likely to be attacked, which greatly affects the performance of the global model

However, the common method of reducing the dimensionality of updated intermediate data to identify malicious users is suitable for non-federated learning frameworks, and cannot play a good role in the scenario of malicious user identification in federated learning.

Images