Re: [babel] Minor clarification to HMAC

[Markus Stenberg <[e-mail]>]

I think I am being dense (I haven’t looked at the protocol in awhile), but 

a) babel-hmac does not have even reference to link-local address, and
b) 6126bis states it is used only to index remote entries

Doesn’t this mean that by forging the link-local addresses (which are not included in HMAC) and then replaying valid HMAC packets you can fill table as you want if this consideration is relaxed?

-Markus

> On 29 Jun 2019, at 13.14, Juliusz Chroboczek <[e-mail]> wrote:
> 
>> Is there some mechanism in place that prevents replay of historic
>> packets? As the HMAC key itself does not change over time, storing
>> (large amounts of) historic payloads with valid HMAC is not hard.
>> Tthe value is bit questionable GIVEN router ID is stable, but if it is
>> not, there is actually potential attack vector here..
> 
> It's not the router ID, it's the link-local IP address.  In order to
> create /n/ neighbour entries, an attacker needs to have captured correctly
> signed packets with /n/ distinct source IPs.  (That's per interface.  If
> an attacker is able to spoof packets on multiple interfaces, then the
> situation becomes worse.)
> 
> (I'm tempted to follow Mahesh's advice, and rewrite the last two
> paragraphs of the Security Considerations section with reference to
> RFC 4987.  Advice?)
> 
> -- Juliusz
>