RE: FW: IPS-All: Reminder - Security draft last call ends Monday, Jul y 1 at 8am EST

[Paul Koning <ni1d@arrl.net>]

Excerpt of message (sent 1 July 2002) by Black_David@emc.com:
> > > 2.3.1.  Transforms "When ESP is utilized, per-packet data origin
> > > authentication, integrity and replay protection MUST be used." 
> > > 
> > > In iSCSI, the replay protection is MUST implement (not MUST use): 
> > > 7.3.1 Data Integrity and Authentication 
> > > "The ESP anti-replay service MUST also be implemented." 
> > > 
> > > (I'm not sure if the security or iSCSI should be changed ? 
> > I think the recent tendency was not to impose IPsec requirements unless
> > they are justified by IPS uniqueness compare to other IPsec usage
> scenarios) 
> > > 
> > > 
> > > +++ I assume security draft will be fixed +++ 
> > 
> > Because of the Bellovin attack on encryption-only ESP, I believe that
> > the first of the two statements is the right one.
> > 
> > There's a lot of argument that integrity should be mandatory in ESP
> > across the board.  The reason why it currently isn't (at least as far
> > as I understand from Steve Kent) is that integrity in the IPsec layer
> > is superfluous if cryptographic integrity is provided at a higher
> > layer.  That case doesn't apply in IPS, so the risk of Bellovin's
> > attack is real.
> 
> Paul - this is only about the anti-replay service, it does not propose
> to change the current iSCSI and IPS Security draft mandates that integrity
> be "mandatory in ESP across the board".  Are you concerned that anti-replay
> should also be mandatory across the board?

Ok, I didn't realize we're only talking about anti-replay.

Technically it's optional separately from integrity.  In practice,
once you have integrity, the sequence checking is trivial, so I don't
really understand why ESP does that.

So while I don't see a specific security hazard from leaving it out, I
also see no good argument for that flexibilty.

     paul