Hyper-Specific Prefixes on the Internet

The widely accepted boundary for IP prefix announcements is usually a /24 prefix in IPv4 and a /48 prefix in IPv6. On this website, we show more specific prefixes than those boundaries (i.e., /25 to /32 IPv4 prefixes and /49 to /128 IPv6 prefixes). We refer to those prefixes as hyper-specific prefixes (HSPs). We show HSPs visible on public RIPE RIS and Routeviews route collector data.

IPv4

IPv6

Paper

Download paper: Hyper-Specific Prefixes: Gotta Enjoy the Little Things in Interdomain Routing

Abstract: Autonomous Systems (ASes) exchange reachability information between each other using BGP the de-facto standard inter-AS routing protocol. While IPv4 (IPv6) routes more specific than /24 (/48) are commonly filtered (and hence not propagated), route collectors still observe many of them. In this work, we take a closer look at those "hyper-specific" prefixes (HSPs). In particular, we analyze their prevalence, use cases, and whether operators use them intentionally or accidentally. While their total number increases over time, most HSPs can only be seen by route collector peers. Nonetheless, some HSPs can be seen constantly throughout an entire year and propagate widely. We find that most HSPs represent (internal) routes to peering infrastructure or are related to address block relocations or blackholing. While hundreds of operators intentionally add HSPs to well-known routing databases, we observe that many HSPs are possibly accidentally leaked routes.

Authors: Khwaja Zubair Sediqi, Lars Prehn and Oliver Gasser.

Download paper Bibtex