Landmark-v6: A stable IPv6 landmark representation method based on multi-feature clustering

Zhaorui Ma, Xinhao Hu, Fenlin Liu, Xiangyang Luo, Shicheng Zhang, Wenxin Tai, Guoming Ren, Zheng Er, Mingming Xu

Abstract: Highly reliable network entity landmarks are crucial for applications like geolocation-aware personalized service recommendations, traceability, and fraud detection. Traditionally, landmark acquisition methods have relied on data mining of rules or network behaviours to establish mappings between IP addresses and geolocation information. However, IPv6 address allocation policies, due to their dynamics and multi-homing phenomenon, pose a risk of IPv6 address deactivation for traditional IPv6 landmarks. To address the issues of reduced numbers and instability in traditional IPv6 landmarks, we propose a novel IPv6 landmark representation method, “landmark-v6”, which is grounded in multi-feature clustering. Firstly, IPv6 addresses are filtered based on multiple attributes derived from network entity fingerprints and routing features. Subsequently, a set of IPv6 addresses is associated with another set through multi-feature clustering. Second, the fine-grained IPv6 addresses are further refined by clustering based on precise physical spatial geolocation information, resulting in candidate landmarks that consist of IPv6 prefixes and geolocation data. Finally, the reliability of these landmarks is determined and evaluated using the voting resolution mechanism in the Candidate Landmark Evaluation task. Our experimental evaluation, spanning 10 months and conducted in three real-world areas, Zhengzhou, Hong Kong, and Shanghai, demonstrates the effectiveness of landmark-v6. Specifically, landmark-v6 obtains 933, 746, and 859 IPv6 prefix landmarks in Zhengzhou, Hong Kong, and Shanghai, respectively. These results surpass those obtained with existing rule or network behaviour-based methods such as Structon, SVMM, and SLG. Landmark-v6 offers a more robust and accurate approach to acquiring IPv6 landmarks, making it well-suited for various applications that necessitate reliable geolocation information. It effectively tackles the challenges posed by the dynamic nature of IPv6 addresses, enhancing both the stability.