Re: [TSCM-L] {1498} Re: Looking for hardware storage device with erase

From: Thomas Shaddack <tsc..._at_shaddack.mauriceward.com>
Date: Mon, 30 Apr 2007 18:55:26 +0200 (CEST)

That very much depends on how you count the cost and if you include the
work on the software. (Maybe there's an open-source application already
available on the Net?) The chips I mentioned are about $2..$4 themselves
and unless we need fancy shielding and more enhanced tamper detection than
a case opening switch, the rest of the parts should be peanuts as well.
However then there is the software and the wiring and supporting circuits
that depend on the details of the application, which I am not clear on.
Making a core that spits out a couple bytes when asked, and forgets them
when told, without side-channel attack protection, is a one-nighter for a
moderately skilled technician.

The question can not be answered well without having more detailed project
specs - what is the threat model, what authorization (if any) the unit
should perform before spitting out the key, what events should cause unit
wiping, what are the impacts of different failure modes (which you can't
prevent but you can pick the direction on which the system will tilt when
pushed - sometimes it is more of a loss if the key is mistakenly erased by
too sensitive trigger system (or a system error) than its leak, sometimes
a failure to erase the key may cause grave dangers and it is better to
mistrigger). I was thinking about this kind of application for quite a
long time already, but did not get enough round tuits yet, nor a suitable
usage scenario, to actually design a proof-of-concept device. Would be
fun, though.



On Mon, 30 Apr 2007, kondrak wrote:

>
> Methinks that would hugely exceed the $300 limit!
>
> Thomas Shaddack wrote:
> >
> > The PCF8570 chip? It is a 256-byte SRAM with I2C interface. Use a small
> > battery for data retention, and cut power when the case is opened. Perhaps
> > use a supercapacitor charged from the power supply and shorted with a
> > switch inside the case when the case is opened (advantage is that the
> > supply voltage goes immediately to ground instead of potentially lingering
> > on a capacitor; the chip itself needs only 1V/400nA (yes, 0.4 microamp)
> > for data retention).
> >
> > Another possibility is using a microcontroller with a suitable program.
> > This choice offers a wide range of options for the device behavior, tamper
> > detection, PIN lock for data retrieval, etc.
> >
> > Choice of an optimal approach strongly depends on the specific demands on
> > the application, the security model, and the balance of losses when the
> > device misfires because it is too paranoid vs when the device does not
> > erase itself because it misses a legitimate trigger event. For the cost of
> > considerably more complicated code, we could also store the data inside
> > the chip in an encrypted state, eg. make the key a hash of a PIN and a
> > stored secret. Or perhaps a multipart secret, where the parts can be
> > fetched over the network, after authorization. Way too many solutions are
> > possible here to pick/suggest any without knowing more about the
> > application.
> >
> > How it should be connected into the computer? Some motherboards have a
> > connector with SMBUS. Is it this case? Or perhaps via USB or a RS232 port
> > or a parallel port connector accessible inside the case?
> >
> > Is it a simple low-stakes application (the $300 budget suggests so), or
> > does the threat model include people with electron microscopes and
> > expensive laboratories, capable of retrieving remanent charge from memory
> > cells, or able to see the changes in their structure caused by prolonged
> > storage of the same value (exotic and limited, but somewhat doable)? Can
> > the adversary monitor the EM emissions of the chip in operation, or can we
> > afford to not bother with those countermeasures in the design?
> >
> > How skilled hardware designer are you? Are you limited to off-the-shelf
> > commercial solutions, or can you build your own if pointed the right way?
> >
> >
> >
> >
> > On Sun, 29 Apr 2007, kondrak wrote:
> >
> >> A cryptostick USB drive and folder lock?
> >>
> >> SynRG wrote:
> >>> We are creating a server appliance that we wish to protect with disk
> >>> encryption. We would like to store the encryption key in some sort of
> >>> volatile / non-volatile storage device that, when the case is opened
> >>> is erased from the storage device.
> >>>
> >>> Obviously we don't need much space, just a few bytes actually. I
> >>> would like to find something for less than $300 if possible.
> >>>
> >>> Do you know of any options?
> >>>
> >>> Thanks.
> >>>
> >>>
> >>
> >
> > >
> >
>
>
> >
Received on Sat Mar 02 2024 - 00:57:17 CST