A Novel Collaborative SRU Network With Dynamic Behaviour Aggregation, Reduced Communication Overhead and Explainable...

Area of Science:

• Biomedical data security
• Healthcare informatics
• Network security

Background:

• Biomedical data leakage and tampering pose significant risks to privacy, security, and reputation of medical networks.
• Existing security models often struggle with dynamic threats and computational efficiency in smart healthcare systems.

Purpose of the Study:

• To propose a novel, privacy-preserving security model for the collection and transmission of biomedical data.
• To enhance the security and efficiency of smart healthcare systems through advanced algorithms and network design.

Main Methods:

• Development of a threat-vector database based on dynamic behaviors of smart healthcare systems.
• Design of an improved, privacy-preserved SRU network to address fading gradient issues and reduce computational cost.
• Deployment of an intelligent federated learning algorithm for collaborative, personalized security modeling without privacy loss.

Main Results:

• The proposed model demonstrates high accuracy in detecting severe security threats.
• Achieved reduction in communication overhead and computational cost through dynamic behavior aggregation and client adjustment.
• Enhanced privacy preservation for biomedical data during collection and transmission.

Conclusions:

• The novel security model offers a computationally effective and parallelizable solution for securing smart healthcare systems.
• The federated learning approach enables collaborative security without compromising individual network privacy.
• The method provides enhanced threat detection, reduced overhead, and improved data privacy compared to existing approaches.