Vertical federal model defense method based on node embedding difference detection

Abstract

The invention discloses a vertical federal model defense method based on node embedding difference detection, which comprises the following steps: (1) a training participant train san edge graph convolution model by using local data and gradient information issued by a server to obtain node embedding vectors updated by the model; (2) the training participant creates a reference map convolution model by using local data and carries out training to obtain a node embedding vector updated by the model; (3) the node similarity between the node embedding vectors updated by the two models is updated, and a similarity difference matrix of the node embedding vectors is updated; (4) the nodes are clustered by taking the similarity difference of the nodes as a node feature, and a target node is screened out; and (5) the node embedding vector uploaded to the server is corrected according to the similarity between the target node and the neighbor node. According to the method, the influence caused by the counterattack carried out by the malicious participant can be effectively weakened, and the robustness of the vertical federal model on the graph data to the counterattack is improved.

Description

technical field [0001] The invention belongs to the technical field of network security, and in particular relates to a vertical federation model defense method based on node embedding difference detection. Background technique [0002] With the rise of deep learning, AI has been widely used in production and life, and a good AI needs a lot of high-quality data for learning. However, in most fields, AI is facing two major dilemmas: (1) data island problem: the data owned by most enterprises is limited or of poor quality; (2) data privacy problem: people's awareness of data protection is also low. It is gradually strengthening, and the management of user data privacy and security is becoming increasingly strict. If the training data is collected and stored in a machine or data center, as the amount of data increases, infrastructure construction needs to be continuously increased. In addition, there are risks of data leakage and data monopoly during the process of data collec...

AI Technical Summary

Problems solved by technology

However, in most fields, AI is facing two major dilemmas: (1) data island problem: the data owned by most enterprises is limited or of poor quality; (2) data privacy problem: people's awareness of data protection is also low. Gradually strengthening, stricter user data privacy and security management

Although this method is simple and effective, it is obviously not applicable in the vertical federation scenario because the data does not leave the local

Images