SAIP: Accurate Detection of Anycast Servers with the Rise of Regional Anycast

IP anycast enables multiple servers to distribute the service load by using a common IP address. Regional anycast is an emerging trend for IP anycast service providers because it allows users to reach geographically close service end-points, improving performance. There has been continuous effort in anycast detection, as anycast service providers usually do not disclose their anycast addresses. However, with the rise of regional anycast, we find that prior methods face severe accuracy challenges due to distance inflation and routing dynamics. In this paper, we propose SAIP, a spoofing-based anycast IP address detection method, to accurately detect anycast servers in the presence of regional anycast servers. SAIP uses IP spoofing to enable cross-catchment communication and identifies different anycast replicas based on indicators such as TCP connection states. Our experimental results on the ground truth dataset show that SAIP achieves 100% recall and accuracy in discovering regional anycast prefixes, while iGreedy only identifies 83.9% of them. And SAIP also achieves a 100 % recall rate for all anycast prefixes, compared to iGreedy's 93.2%. Through Internet-wide measurements, SAIP discovers 13,465 anycast IP/24 prefixes, 11.5% more than the existing largest anycast prefix dataset. We provide public access to this data at https://ki3.org.cn/#/datasetDetail?dataset=anycas.