aSparkle

Utility to calculate the global ASPA deployment status at a given RIB snapshot hour. Fully written with ❤️ in 🦀.

Built atop the amazing BGPKit framework and the RPKIViews repository.

Idea

The aSparkle utility compares the ASPA records within a set of .asa files against the routes visible from RIPE RIS and Routeviews. Unlike actual on-path ASPA validators, aSparkle does not have accurate information about the relationships of the ASes that originally observed the routes, i.e., the RIPE RIS and Routeviews vantage points. As this information is neccessary to accurately perform ASPA valiadation, aSparkle opportunistically checks the ASPA state of AS links whenever it has high confidence in its ability to accurately determine a paths up and downstream.

Up- and Downstream Inference.

The ASPA Verification RFC specifies that "The upstream verification algorithm ... is applied when a route is received from a customer or lateral peer, or is received by an RS from an RS-client, or is received by an RS-client from an RS." aSparkle looks at BGP paths collected from hundreds of "randomly" chosen ASes, i.e., the data we receive contains paths that may contain upstreams and downstreams. hence, we first determine each path's up and downstream section via the following set of inference heuristics:

1. If we encounter one of the well-known provider-free network (also known as Tier 1 ASNs), we assume that every hop between the Origin and the Tier 1 belongs to the upstream. If we find that the ASN after the Tier 1 is a non-transparent route server, we infer the route server to be part of the upstream (as CAS are supposed to include non-transparent route servers within their aspa records.)
2. If we encounter a non-transparent route server, we assume that every hop between the origin and the route server is part of the upstream.
3. if we encounter an hop with an valid aspa attestation, we assume that all hops between the origin and the CAS's provider are part of the upstream. Notably, if we find multiple valid ASPA's, we choose the one furthest away from the origin.
4. If we receive data from a route collector peer that is a route server, we assume that all its paths consist of only upstreams.
5. BGP Roles.