Data Access Security Protection Methods for Dance Movement Therapy Combined with Image Analysis

Abstract

In DMT, data such as dance videos and facial expressions provide important emotional clues. Traditional data protection methods cannot effectively protect the privacy of image data and are prone to leakage. This paper introduces a new data security protection method by combining the GraphSAGE algorithm and image analysis technology, aiming to strengthen the privacy protection of image data and optimize the access control of treatment data. First, by converting treatment links, participants, emotional states, etc., into graph nodes, defining edge relationships, and using GraphSAGE for node feature learning, a graph representation of the patient's emotions and treatment process is constructed. Secondly, CNN (Convolutional Neural Network) is used to extract emotional and action features from dance videos, and then fused with text and behavior data to generate multimodal feature vectors, which are further input into the GraphSAGE model for learning. To protect privacy, encrypted GNN is used to encrypt data, and a dynamic access control strategy is designed based on user roles and data sensitivity to ensure that all parties only access authorized data. Finally, through sentiment analysis and GraphSAGE model, combined with the automated risk alarm mechanism, abnormal behaviors or emotional fluctuations during treatment can be monitored and responded to promptly, thus improving the safety and effectiveness of treatment. Experiments show that the encrypted data leakage rate is stable at 0.01, and the user access success rate is above 90%. This method provides a new idea for data privacy protection in the DMT field.