Summary

  • Dan Li, Lancheng Qin, Li Chen and Libin Liu propose Bicone SAV in an individual Internet-Draft dated 17 August 2026. It addresses customer and lateral-peer AS interfaces whose source-prefix allowlists may be incomplete.
  • The draft gives three dispositions: permit an allowlisted source, drop a source positively identified on a separate blocklist, and permit an unverified source while logging or monitoring it. A packet missing from an incomplete list has not earned a guilty verdict.

The missing name is not a conviction

The first list says who belongs. Operators can derive it from a neighbouring AS's customer cone, then reject source addresses they do not see. That is quick to explain at a change meeting. It is harder to defend when a legitimate prefix never propagated far enough to enter the list. The Bicone SAV draft gives examples involving limited propagation and a service address used in Direct Server Return. In each, the packet can be genuine while the allowlist is silent.

Dan Li's team keeps the useful half of the rule. A match on the allowlist still admits the packet. For a discard, it asks for affirmative evidence on a second list. The proposed blocklist draws primarily on prefixes tied exclusively to the provider cone: sources that should not appear on a customer or lateral-peer ingress. The draft uses ASPA and AS_PATH information to build the cone, and ROA or proposed TOA records to connect prefixes to it. Prefixes that might legitimately belong outside that cone are removed from the candidate set.

That leaves a third bucket. The source is on neither list; the packet passes, and the operator records or monitors it. This is the uncomfortable part of a security control: a spoofed source may pass through while the evidence is incomplete. It is also the part that prevents an incomplete inventory from silently cutting off legitimate traffic. The draft offers a decision order, not a measured deployment result.

When the lists quarrel

The proposal also specifies a revealing collision. If both lists cover the same address, the draft says to expose the contradiction for investigation and avoid an automatic drop based solely on the conflicting information. The filter is no longer allowed to disguise a data disagreement as certainty.

Partial ASPA, ROA or TOA coverage makes the blocklist smaller. It does not, in the authors' design, enlarge the permission to discard. That trade is the real wager: accept a monitored unknown today and demand better authority data tomorrow. RFC 8704 provides the earlier feasible-path filtering context; Bicone remains an individual proposal, with no published production false-block or false-permit count. Such a count, tied to the list version and ingress interface, would tell operators whether the third door earns its place.

Dan Li's Tsinghua University profile places him at the NASP networking lab. Here the researcher's contribution is a small refusal with a large operating consequence: missing evidence gets its own state. It does not get promoted into an accusation by a one-line filter.

Sources: Bicone SAV individual draft and status; draft revision 01; RFC 8704; Tsinghua University profile.