Re: [dhcwg] more thoughts about draft-ietf-dhc-sedhcpv6-02.txt

[Sten Carlsen <[e-mail]>]

> On 26 Jun 2014, at 19:28, Francis Dupont <[e-mail]> wrote:
> 
> In your previous mail you wrote:
> 
>>> => I can't see why you say that or do you mean timestamps have another
>>> role than anti-replay?
>> My point is that giving the attacker the means to synchronise the
>> replay attack with the target will give replay attacks a considerable
>> time window in which to play.
> 
> => synchronization allows to offer a smaller window.
On the contrary. Synchronisation allows me(the attacker) to change the victim's time to suit my replay.
> I have two other arguments:
> - IMHO it is better to consider clients will be synchronize than
>  to allow a large window and expect attackers to be never synchronized
Attackers will synchronise but to their liking, in effect the attacker will always be synchronised.
> - it is clear timestamps by themselves are not enough, this is why
>  I proposed to add nonces (note transaction IDs already provides
>  24 bit of unpredictable data but 24 bits are not enough for high
>  speed LANs)
> 
>> If signing messages in itself protects against replays, why bother
>> with synchronising time?
> 
> => signing is for origin and integrity, timestamps and/or nonces
> are for anti-replay. Note it seems we need more analysis about
> the usefulness of the anti-replay (I thought about the obvious
> RELEASE replayed after a long delay but anyway a good security
> threat analysis is (*again*) hardly needed... argh!)
> 
> Regards
> 
> [e-mail]