This is done by inserting into the
normal instructions the processor should be receiving, bogus instructions that
lead to the processor erroneously spewing out information it should not.
Why doesn’t that
all happen with a trusted core? Actually, it could, and that’s why there are
several layers of defense inside a secure core. They are all based on
establishing trust in both the processes of building the secure core and in its
physical integrity. The programs that are loaded in a secure core at the
factory have been validated and verified using standards of security. When
there is a need to load a program that has not been so validated and verified,
it is not allowed to directly instruct the processor. It can only talk to an intermediate
interpreter that checks that each instruction is acceptable, and not one that
threatens the integrity of processing. Since the trusted core is physically
protected, once all this apparatus in on board, there is no easy way to physically
affect it. By controlling with equal strength the sensori-motor elements of the
secure core, the critical keys are protected and the digital signature can be
trusted as being unique to the personal electronic device containing the
trusted core.
Now comes the
deception. If my personal electronic device does not have a secure core, and
consequently its secret key is stolen by another computer as we have seen is
possible, then this other computer can masquerade as my device, which as we
also know is supposed to represent me on the network. When the digital
signature of my personal electronic device is copied, it is, in the very legal
sense, my signature which is compromised. When the offending computer now
starts to use my digital signature to purchase goods or to sign tax documents,
or to perform any other electronic transaction of importance, the level of
trust assigned by the institution receiving the information is very high if
that institution believes that the key used for the signature was protected.
Trust extends by causality; it conveys deception as well as it conveys honest
credentials. What went wrong? As we suggested earlier, trust applies
recursively in the full causality chain. This means it applies also to
causality itself. The receiving computer should question if causality is
warranted, which means it should question whether the original trust is
warranted.
Subsequently,
the proper way for an institution to accept a digital signature is to first get
assurance that it comes from a trusted core. This is done by first checking the
digital signature of the trusted core itself. Or course, to enable this, there
must be a trusted core in my personal electronic device. As you may be
wondering how we’d know that the digital signature of a trusted core is not
itself compromised, we need to explain that part of the intricate security
procedures of the trusted core is the way to load an identification key. The
proper way to do it is to let the secure core itself generate its own keys, so
that they never appear anywhere outside of the secure core. The way it is done
is a little bit too involved for this discussion, but we can observe that it
has to do with the mathematics of large prime number generation. Now, while
trust can be imbued to secure cores, it can not be imbued to a personal
electronic device without a secure core. So trusting a non protected personal
electronic device without checking that it has a secure core is more
threatening than would initially be considered. Once trust is assigned to the
wrong device, it can extend in many directions, thus propagating the deception
to great length. If we trust the priest, then the priest can abuse that trust.
So we have to be sure that we select our priests carefully. It’s the same for
personal electronic devices.
As we have said,
trust is recursive. Trust applies to trust, and therefore we would expect
deception to apply to deception. For example, let us consider a personal
electronic device with a secure core, regularly doing transactions on the Web
using its digital signature facility. We can consider someone who, observing so
many successful transactions might make the natural assumption that commercial
and other entities trust the personal electronic device because it has a secure
core.
|