[Linux] Re: Decrypting a pointsec HDD

[linux@flux.org]

On Sat, Feb 23, 2008 at 12:03 PM, bob daggit - daggit_2@hotmail.com
<+flux+simicich+5182c56311.daggit_2#hotmail.com@spamgourmet.com>
wrote:
>
> just in case this also applies, a scathing expose of how to cold-boot a computer and retrieve the hdd encryption key from the RAM is at
> http://www.linuxtoday.com/security/2008022201626SCSW
> and also
> http://lwn.net/Articles/270314/
>
> it makes one wary of some hdd encryption systems.

There is really nothing new here.  Think about it.  When you use a
computer to decrypt something, how do you do it?  You enter the
decryption key from the keyboard or some other entry device.  Say you
have a fingerprint reader that algorithmically converts your
fingerprint to a key that it uses to encrypt or decrypt the data.

Is this secure?  Even assuming your fingerprint can't be reproduced
(which has long be debunked - heck, mythbusters did it and even used a
technique that puts the methodology into pretty much anyone's reach)
all that is going to happen is that some software is going to convert
something (fingerprint, typed or spoken passphrase) into a number that
will eventually be used as a key.

Intercept the key, and you don't need the original phassphrase or fingerprint.

If someone can install programs on the computer, it is not secure.

If you can take the hardware apart, it is not secure. That is not new,
this just is another amplification of that principle.

This was a demonstration by researchers that they could recover the
contents of DRAM after the computer was turned off - and that if you
got the chips cold that the delay before recovery could be hours.

The security of a computer depends on the physical security of the
computer - as well as the security of the people around it.  You need
a plan B if your location is compromised. This is why the loss of the
USS Pueblo was such a blow. When I was in the Navy, card files that
contained top secret data were generally wired with thermite systems
so that if you had to abandon ship you had a way tto make sure that
they were not captured. The Captain of the Pueblo asked for a destruct
system - and got a 55 gallon drum welded to the ship's fantail so that
confidential trash could be disposed of - but nothing that would
handle codebooks and so forth.

My point is that the problem long predates computers. Physical things
need physical security.

There have been attempts to overcome insecure remote computers.
AFS/Kerberos was one of them:  How do you protect computer users where
users are infinitely intelligent and perhaps bored (MIT students) from
each other - one person uses the computer and then the next person
does.  Assume the first user cracked the computer - how can you be
sure that the second computer user is not typing their login into a
trojan?

The only way they could be sure to provide user to user isolation was
to reinstall between users.

But even then, there was a presumption that the person in front of the
computer owned the data that their credentials claimed that they
owned.  The problem that someone unauthorized may be accessing the
data is a much harder problem.  Implanted RFID chips?

Ram holds its contents.  In the old days, when magnetic core was the
standard, programs would sometimes fail randomly - because they would
not clear their memory contents and the leftovers might be valid
numbers of they might not.  Sometimes from the day - or years before.

Then there is the whole paging file issue - gpg needs to be installed
setuid root - so that it can page fix important pages - that hold
decrypted keys - so that they don't get paged out. But any other root
program can look into that memory and get the decrypted keys.

I remember working in, what at the time. was said to be a highly
secure commercial environment - where people had tight control over
disk contents, and you needed a special security allowance to get a
disk allocation and when disk was wiped between projects, it was
always reformatted (in the old days, this was not hard) and I recall
being able to dump their highly confidential files without any problem
- because they only cleared the first few inches of a tape to write a
new label and a couple tape marks when reusing them, so the rest of
the tape could be read by the new user using by simply telling the
sequential file access method to ignore I/O errors and to keep reading
even though end of file had been reached.  It was "too expensive" to
erase tapes between users - until I wrote a program that kept track of
who had the tape last and instead of calling for a scratch would call
for one of the tapes that they were the last user of from the scratch
pool - and which read the label and if someone else's scratch was
mounted instead of the one called for and it was not owned by that
user would erase it before turning control over to the user.

Of course, people like the CIA were more secure than we were.  They
had a rule - no disks ever left their premesis. At the time, disks
would be rebuilt - if you had a crash, they would replace the damaged
platters, reformat, and put the disks back into service, so the broken
disks were valuable.  But the CIA would not allow IBM to remove the
disks, they would only allow new ones to be brought in, and when IBM
wanted to charge them the core charge they just paid it - and melted
the drives down, with highly paid data security experts supervising.
I had friends who worked in equally secure places, and they burned the
chad from punch cards, under the supervision of senior enlisted types
- an E-8 Chief Petty Officer had to go to the dump and supervise the
E-4's as they raked through the chad to insure it all burned to ash
and that the ash was stirred so that it could not be reconstituted and
that there was nothing hidden in it.

These measures seemed unreasonable to at least some people, but they
were specific applications of general rules that are very valid.

Anyway, looking at the referenced articles led me to some funny stuff.
 You are a computer tech working for a police department.  Your guys
just served a no-knock warrant and caught the bookies before they
could hit the panic button - all the computers are ready to go - and
the hard drives are decrypted - you could copy all those incriminating
spreadsheets - so long as the computers stay powered up and the screen
saver (which re-encrypts all the files, and to which you don't know
the password) does not come on. The people working there won't help
you - if they do, they get a public defender instead of a good mob
attorney, and if you don't get the spreadsheets you have no evidence.

So what do you have that helps you move the boxes without letting them
become uncomprimisable - so that the evidence is still available to
your forensic analysists? (Remember that at this point in time, the
code makers are way ahead of the code breakers - if you have picked a
good password and the password is not available to the decoder, the
data probably can't be touched, despite NCIS and Num3ers and the rest
of the pseudo-science shows.)

One item is something that you can plug into the power strip - or the
other outlet of a duplex outlet - or even clip into the power cord
itself after dissecting it - which will plug into a UPS and provide
power when the mains are disconnected.  So now you can move the comp
from the rat infested bookie joint to your lab.

The other device is something called a mouse jiggler - which plugs
into a USB port and looks like a second mouse. It sends constant
motions to the system, which keeps the screen saver from coming on.

Hmmm...This calls for a program that runs constantly and (1) causes
the screen saver to come on when the mouse has been moved for 20
minutes and no keyboard input has been received and (2) an outlet
which looks completely normal but which cuts power from side (a) when
either something is physically plugged in to side (b) or the outlet is
dislodged in its box.  Might be good enough to short everything to
make the breaker trip - so that all the computers on that circuit
drop.

And. of course, if the program detects a second mouse being plugged in
(or for that matter if any device reconfiguration is done) it does the
encryption thing.

I'm experienced in these matters but I do not think that this is
something that a bad guy would not think of if this was the sort of
data that needed to be protected.  These people have already gone
through the trouble of installing encryption, and they have already
set up the sort of encryption that reinstates itself on power down or
disuse.  The good guys are doing technological things so that they can
move computers.

Do you expect the bad guys not to do more technological things so that
the computers can't be moved?  Mercury switch on the low voltage
circuit to cause internal surges in case of tilt, movement or
vibration?  Microswitch that shorts the power if the case is opened?

Most case power intrusion detection circuits are designed for a
machine owner to detect that a person who works for them has opened
the computer case - they put up a message on every reboot until reset
with a power on password.

But the simple answer to any of these things is just to require a
password be entered every 30 minutes.

And to disable the USB ports.  Why would anyone make a secure computer
that someone can plug random devices into?  A few drops of superglue
to permanently attach the keyboard and mouse - and a few more drops of
superglue to disable the rest of the ports, oor things that look like
USB ports, but which shut the computers down when you plug something
into them.

So, case protection, intrusion detection programs that don't do what
the good guys expect, and a limited UI.  No way to plug in a computer
hard drive or a communications device without the computer noticing.
All actual comm done by modem with the channel secured by something
that uses a handshake like TLS - but which does not use a generic PPP
or SL/IP sort of thing. So you know who you are talking to but the
channel can't be subverted for a side purpose.

I guess I don't see it as that hard for a bad guy to make a computer
that could not be dissected to any avail by a forensic guy.  I know
the stuff they teach in the courses and I know the sort of person who
takes them. They are well meaning people who, well, really are
excellent cops and who are not the sort of person you'd hire to
administer your system.

If they have the right tools, they will at least be able to preserve
the evidence - but those tools, and the procedures they follow to
preserve the evidence, actually stop them from being able to pull off
something like the jiggler. :-)

I've seen gadgets that people could wear that automatically activated
their computer as they sat at it - and deactivated it when they walked
away.  Combine that with manual password activation, and as soon as
you cuff the guy and drag him away from the comp, the evidence hides
itself.  It all depends on what the bad guys think it is worth to make
something secure - and these are just ordinary non-violent criminals
we are talking about - a really bad guy, you'd think, would be able to
do something far more effective than I could dream up in a few minutes
of musing.
-- 
A man can't live in the everglades
Where a man can hide and never be found and have no fear of the bayin' hound
But he better keep movin' and don't stand still
If the skeeters don't get him then the gators will