Exploring the Temporal Degradation and Drift of AS Path Inference

The Internet inter-domain paths, i.e., the AS paths, are important for network management, traffic engineering, and security. Due to business confidentiality, security, and privacy, the AS path information is non-public. Due to limited measurement resources, obtaining AS path information by measurement-based approaches is not scalable. Therefore, path inference approaches are proposed to broaden the availability of path information. These approaches assume that AS paths remain stable over a certain period of time, yet conflicting research findings question this assumption. Furthermore, the duration of the “certain period of time” is not clearly defined. Thus, we aim to address the following question: “How do the performance and temporal drift of path inference approaches evolve over time?” In this paper, we conduct a quantitative validation study and a temporal drift analysis to examine the evolution of AS path inference performance over time. The quantitative validation study shows that the minimal performance degradation is only 2.09% over eight weeks. The temporal drift analysis shows that, among the three evaluated methods, KnownPath exhibits the slowest drift, GMPI shows a moderate drift rate, and ProbInfer drifts the fastest under the current decision rule. The results provide preliminary evidence on how historical data can be leveraged despite limited measurement resources and can inform refresh-frequency decisions for path inference services under computational constraints.