GNN-based Method for Predicting Access Control Relationships for Big Data

Access control is one of the important technical means to achieve secure and controllable big data, but the lack of relational data inevitably affects the completeness of relational-based access control rules. Although existing link prediction techniques can be used for the prediction of entity relationships, they still face problems such as incomplete information elements and undifferentiated entity weights. To address these problems, this paper proposes a GNN-based link prediction model LPMG. The model proposes a node embedding feature learning method based on directed neighbor subgraphs, incorporates elements such as attention coefficients and relationship types, improves the representation capability of node embedding, and achieves link prediction of entity relationship graphs. The experiment results show that the model effectively improves the effect of link prediction and can meet the demand of big data access control relationship prediction.