Research on enterprise data sovereignty security control systems with intelligent detection and recognition of multimodal data

The introduction of intelligent algorithms drives the intelligent development of enterprise data sovereignty management. A multimodal detection model integrating text, image, and behavioral features is constructed. By embedding an attention-weighted mechanism within the Bi-GRU architecture, it achieves adaptive fusion of cross-modal features and temporal aggregation, enhancing the ability to identify unauthorized access and abnormal operations. Experiments demonstrate that this model achieves 4.7% higher accuracy, 2.3% lower false positive rate, and approximately 18% shorter response time compared to traditional RNNs on datasets of equivalent scale, delivering high detection precision and real-time performance.