> For the host to which the attack packets are destined, I don't think that 
> LISP is worse in this case. However, it is not the effect on a host that I am 
> concerned about. If the xTR serving that host in this case is either a CE 
> router or a PE router, then this is a lot worse for the xTR, since the attack 
> packets cause a reaction in its control plane. Of course if the effect of the 
> attack is to fill the cache, then again LISP is worse since "non-LISP 
> Internet operation" doesn't have a cache. I understand that we are not 
> planning to describe solutions in this text, but if rate limiting of map 
> requests is considered to be normal practice, I wonder if it would be worth 
> mentioning that rate limiting could prevent overload of the xTR's control 
> plane, but that in this case the attack could prevent map requests in support 
> of legitimate users. 

I beg to differ, because what you refer to a cache makes assumptions that it is 
finite size. A BGP RIB is finite size as well. And even a legit BGP peer can 
fill up a finite size RIB.

So it is my claim that LISP is no worse.

> the attacked xTR noticing and putting in a packet filter for that source RLOC,
>>> so that varying the source RLOC may make this attack more difficult to 
>>> counter. 
>> 
>> You shouldn't say how the xTR would handle this or else, yes, it will get 
>> worse.
>> Because if you choose a faulty solution, you will introduce more issues.
> 
> I am fine with dropping the sentence stating "Optionally the attacker could 
> use the same source RLOC for each, but this could in theory lead to the 
> attacked xTR noticing and putting in a packet filter for that source RLOC, 

I wouldn't put in a packet filter or else the real RLOC is denied service by 
the attacked xTR itself.

> so that varying the source RLOC may make this attack more difficult to 
> counter". However, I do wonder, what would happen if a single attacker sent a 
> large number of packets which look like LISP packets with the same RLOC for 
> all (perhaps the legitimate RLOC of the attacker) but with different EID's, 
> each EID chosen from a different EID block. Would the xTR receiving all of 
> these incoming packets send a map request for each EID? 

This is what Ron brought up.

> This could be a way to get around source IP address checks (where they are 
> deployed), since the source RLOC would be correct and legitimate. 

Right, understand.

It all comes down to the (source-EID, source-RLOC)-tuple needs to be 
authenticated to be the legitimate source.

Dino



_______________________________________________
lisp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lisp

Reply via email to