Balancing Interests: A Policy-Based Approach to Internet Peer Selection

This paper presents a novel, policy-aware approach to selecting and evaluating potential peering partners among Autonomous Systems in the Internet ecosystem. We develop a quantitative scoring model that combines requirements from peering policy documents with AS profile features from public datasets. Our model computes compatibility scores for potential AS peering pairs based on mutual requirement satisfaction, addressing limitations in existing approaches that often overlook specific policy requirements. We validate the proposed model using machine learning techniques and demonstrate its robust predictive performance across diverse ASes. The results reveal the individualized nature of peering strategies and the varying importance of different factors across ASes. This research contributes to a deeper understanding of Internet interconnection dynamics and provides a valuable tool for network operators to optimize their peering strategies.