New Privacy-Preserving Routing Method Cuts Multi-Domain Path Calculation Time by 24%, Keeps 90% Bandwidth Efficiency | Newswise

New Privacy-Preserving Routing Method Cuts Multi-Domain Path Calculation Time by 24%, Keeps 90% Bandwidth Efficiency | Newswise

Researchers have developed BI-TE, a novel algorithm that resolves the long-standing conflict between network privacy and performance in multi-domain environments. By integrating differential privacy with Graph Neural Networks, the system allows distinct administrative entities to collaborate on traffic routing without exposing sensitive internal topologies or bandwidth data. This innovation demonstrates that organizations can achieve significant efficiency gains while strictly maintaining the confidentiality of their proprietary network information, effectively breaking the barrier that previously hindered cross-domain cooperation. The core mechanism involves local controllers adding controlled noise to their network reports before sharing them, ensuring that individual domain secrets remain indistinguishable to external parties. A central coordinator uses a trained neural model to accurately predict actual bandwidth usage from these perturbed inputs, enabling near-optimal path selection. This approach eliminates the need for raw data exchange, proving that sophisticated machine learning can compensate for the inaccuracies introduced by privacy-preserving techniques, thereby maintaining high link utilization and fair traffic distribution. This development is highly relevant to the open data movement, as it provides a technical framework for secure data sharing among independent entities. It illustrates how privacy-enhancing technologies can enable collaborative ecosystems without compromising individual autonomy or intellectual property. By making inter-domain collaboration feasible and safe, BI-TE offers a scalable model for other industries seeking to leverage collective data insights while adhering to strict confidentiality standards, bridging the gap between openness and security.

Source: newswise.com
Published on 2025-07-26