Using GPG in embedded applications?
bo.berglund at agiusa.com
Thu Mar 26 10:54:52 CET 2009
The threat mode is concerning the hacking of PC based software.
We want to place a vital part of the processing of data on a unit consisting
of a microcontroller device which we design and build ourselves. It is not
possioble for a hacker to actually intercept the processing done on board
this unit, but he would probably be able to hook into the transfer of data
between the PC and the unit (USB or serial).
So we want to encrypt the data stream.
And of course the PC software can simply be attached to a debugger and
So now if there is a key pair used with a secret key embedded in the
microcontroller code and the public key in the PC software the PC can
encrypt the data set before transferring to the device using the device's
Then the device can decode it and then perform its processing after which it
will encrypt it using some key before sending the data back.
By using its own secret key to encode the result it would be possible to
decode with the public key and the PC would get the processed data back for
But by using a key pair also for the PC it would be even more difficult to
break, because then the device would encrypt using the public key of the PC
when sending back and the PC would decrypt using its "secret" key. Of course
the PC secret key would not be protected against debugger tracing, but since
the debugger cannot see the public key in the device it cannot find out how
the data should be formatted in a pirated device to mimick our device.
Should work to deter hacking, I believe.
But the problem is how to implement encrypting/decrypting on the
From: gnupg-users-bounces at gnupg.org [mailto:gnupg-users-bounces at gnupg.org]
On Behalf Of Nik N
Sent: Thursday, March 26, 2009 6:54 AM
To: Gnupg-users at gnupg.org
Subject: Re: Using GPG in embedded applications?
On Tue, Mar 17, 2009 at 12:24 PM, Bo Berglund <bo.berglund at agiusa.com>
> The idea is to have the PC program encrypt a fairly large chunk of data
using the embedded unit's public key and then send the result over the
channel into the embedded application.
> Inside this (protected) hardware the secret key would be used to decode
the data, then some processing would be done whereupon the resulting data is
again GPG encrypted now with the public key of the PC program and sent back
over the channel.
> Finally the PC program would decode the data and further process it.
I am trying to understand your threat model:
If the attacker has access only to the channel but not to the two
communicating devices, a simpler, symmetrical-cipher-only solution
would suffice. If, on the other hand, the attacher has access to
either device, isn't it reasonable to assume he'd be able to pry the
decryption (private) key and decrypt the data (flowing in at least one
Gnupg-users mailing list
Gnupg-users at gnupg.org
More information about the Gnupg-users