From: "Mark O. Aldrich" <maldrich@grci.com>
Date: Tue, 16 Jul 1996 14:01:11 +0800
To: Deranged Mutant <WlkngOwl@unix.asb.com>
Subject: Re: Opiated file systems
In-Reply-To: <199607151158.HAA28540@unix.asb.com>
Message-ID: <Pine.SCO.3.93.960715172915.7563D-100000@grctechs.va.grci.com>
MIME-Version: 1.0
Content-Type: text/plain


On Mon, 15 Jul 1996, Deranged Mutant wrote:

<snip - much stuff on crypto device drivers for disk subsystems deleted>
> 
> > - Facility for duress key, with the real data hidden in the unused
> >   space of the first encrypted drive.  To increase the plausible
> 
> Huh?!?
> 

Hey, DM, don't laugh.  I've gotten such requests before about crypto
subsystems, including tokens with "protected" keys onboard.  The idea is
that there's a "duress key" or a "panic key" that, when entered, fools
someone into thinking the process is working but, in fact, it's not
working at all and it usually is doing something else (like scrubbing the
hard disk, scrubbing the key PROM, or calling the police).

I've worked at sites that have their electronic door locks rigged the same
say.  The way it works is, let's say a terrorist has a gun to your head
and demands, "let me in the door or I'll blow your head off."  Naturally,
the Government doesn't want you to have to choose between dying and giving
out the cypher lock combination (guess which one people choose in blind
testing?), so you put in the "duress code."  The door unlocks so the
terrorist thinks that all is well.  However, the alarm just went off over
at the security substation and, in about two minutes, a heavily armed SWAT
team will be arriving. 

Same for cypto keys, but with a different "payload" if the duress key is
used.  The data either gets "nuked" (Gosh, Mr. FBI Agent, I *thought* that
was the right crypto key - sorry about destroying the hard disk), the keys
disappear (damn! my fortezza card just zeroized again!), or the data
appears to "decrypt" but it's actually phoney data that's been hidden
somewhere or is 'hard-coded' into the program handling the duress key. 

The payload of getting false data out of a crypto algorithm, such that the
data looks "real", when a duress key is input to the algorithm is not
something that I've seen approached in any reasonable manner.  Probably
because it's just too damn hard and the notion of "real looking" data is a
little hard to define scientifically.  A combination stego/crypto solution
may be more appropriate, but close examination of the box is going to
reveal what happened (assuming the desired solution must withstand some
protracted forensics?).  The nuke_the_data or nuke_the_keys solutions are
easier to do, and have been implemented in several situations of which I
am aware.  

------------------------------------------------------------------------- 
|Just as the strength of the Internet is  |Mark Aldrich                 |
|chaos, so the strength of our liberty    |GRCI INFOSEC Engineering     |
|depends upon the chaos and cacophony of  |maldrich@grci.com            |
|the unfettered speech the First Amendment|MAldrich@dockmaster.ncsc.mil |
|protects - District Judge Stewart Dalzell|                             |
|_______________________________________________________________________|
|The author is PGP Empowered.  Public key at:  finger maldrich@grci.com |
|    The opinions expressed herein are strictly those of the author     |
|         and my employer gets no credit for them whatsoever.           |
-------------------------------------------------------------------------