Recently Jozef Pivarník and Matěj Grégr published an excellent write-up on RA Guard & evasion techniques. Amongst others they tested the “undetermined-transport” ACL we described here and here. As it turns out the “workaround” for implementing undetermined-transport on platforms seemingly not supporting it, causes some bad collateral damage: the respective port does not forward any IPv6 packets any more (this was brought to my attention by Roberto Taccon). We had done some tests after applying it (by means of the “workaround”) but we had just looked at fragmented RA packets (which did not get through => test succeeded). So, frankly: the undetermined-transport trick does not make sense at all on the “unsupported platforms”…
Jim Small didn’t notice this either, in his great presentation at the North American IPv6 Summit (which, btw, to the best of our knowledge is the best overview of ACL approaches to counter common IPv6 attacks on the local link).
Furthermore it should be noted that Jozef and Matej describe some really interesting ways to evade current implementations, incl. an evasion variant merely based on extension headers (without fragmentation) that we hadn’t been aware of before. These will be included in these workshops.
Obviously much more research (and vendor scrutiny) is needed as for RA Guard…
have a great week everybody
Enno