Re: [IPsec] Regarding ISAKMP SA lifetime negotiation.

["Anoop V A (anova)" <[e-mail]>]

Hi Yoav/Kivinen

    Thanks a lot for your quick response.  As  Yoav said - this is for the  some legacy solution support - there in some strange situation we are ending up with different lifetime in the boxes, and creating some issues.
So now as to add to the same question -

 If we are sending the extra IKE responder life time notification in MM6 or AM2 - since the peer AUTHENTICATION data is also available in these messages - can we overcome this situation. We can avoid attack by changing the life time only if the authentication is successful.

I understand may be other implementation will avoid this extra notify - but is there any violation in sending this extra notify in these messages?

Thanks
Anoop

-----Original Message-----
From: Yoav Nir [mailto:[e-mail]] 
Sent: Monday, March 04, 2013 9:30 PM
To: Tero Kivinen
Cc: Anoop V A (anova); [e-mail]
Subject: Re: [IPsec] Regarding ISAKMP SA lifetime negotiation.


On Mar 4, 2013, at 4:31 PM, Tero Kivinen <[e-mail]> wrote:

> Anoop V A (anova) writes:
>> 
>> Hello experts,
>> 
>>   I have a generic doubt regarding the ISAKMP SA(phase 1) life time
>>   negotiation. My  query is can we agree up on the  ISAKMP life
>>   time in the first two messages of MM or AM. 
>> 
>> What I want to know is  - the life time is sent as an proposal 
>> attribute in the first two messages of Main mode and aggressive mode. 
>> We are not negotiating the parameter so if the responder is having a 
>> less life time value configured - then can we transfer this info in 
>> the MM2 or AM2 message from the responder along with the negotiated 
>> proposal attributes. Basically I am trying to change the life time 
>> attribute sent by the initiator - in this scenario.
> 
> Anything extra (notifications etc) you send inside the main mode or 
> agressive mode packets are not authenticated, so sending responder 
> life time notifications is not good idea (and the other end will 
> simply ignore it).

This is true for MM2, but not for MM6. MM6 is encrypted and authenticated, so the peer can and should (if they implemented the draft) use it.

> 
>> We have the responder life time notify mechanism as per the draft 
>> (draft-ietf-ipsec-ike-lifetime-00), but the separate notify messages 
>> are not reliable in IKEv1(Uni directional)
> 
> That is expired very old draft that did not get forward, there is no 
> point of implementing that...

It makes sense for working with implementations where you can't configure such parameters, like VPN clients. Sadly, none of the generic IKEv1-using clients (like the L2TP clients) seem to support this draft.

>> In short my questions are:
>> 
>> 1.       Can we send the responder life time notification in MM6 or
>>         AM2 message from the responder?
> 
> You can, but most likely all implementations will ignore them, and if 
> any implementation does not ignore them, that opens attack where 
> attacker can shorten the lifetime at will by just adding such 
> notification.

MM6 and AM2 are protected.

> 
>> 2.       Or can we alter the life time attribute of the ISAKMP SA
>>         proposal offer?( Is this considers as  a violation of the
>>         RFC)
> 
> That is considered violation of the RFC, thus all complient 
> implemnetations will reject the proposal. RFC2408 section 4.2 Security 
> Association Establishment last paragraph:
> 
> 								The
>   initiator MUST verify that the Security Association payload received
>   from the responder matches one of the proposals sent initially.
> 
> Easy answer to all of your questions is to NOT use the protocol that 
> was obsoleted more than 7 years ago (IKEv1 was obsoleted by IKEv2 in 
> 2005), and instead start to use current version of the protocol i.e 
> IKEv2, which already solves those problem (the notifications are 
> authenticated, the lifetimes are removed and moved to local matter, 
> the informational exchanges are reliable etc).

Now that I agree with. This draft addresses one deficiency in IKEv1. The reason to support IKEv1 today is to support legacy implementations. I don't think anyone is adding this feature now to legacy implementations. So yes, <hat type="vendor"> Check Point code supports it, but </hat> it's better to use IKEv2 where interoperability with different configured lifetimes has been shown to work.

Yoav