Re: Smart Card

Hi,

I'd like to place my 2c worth of opinion into this discussion. What I
have to say is more about the inner working of smart-cards and smart
card readers and a little know type of attacked called power cryptography.

A few years ago I did a small internal research at the place I was
working, on certain exploits that could be carried out on smart cards
and smart card readers.

Now what Peter said about compromised hosts is very true, in fact
there is currently no way of getting around the problem, even the pin
number entry scheme fails in due course due to the prevailing need for
the smart card user wanting to have themselves authenticated on that
particular system.


My fiddling with smart cards centered around a simple gemplus smart
card reader and standard ISO-7816 c-type smart cards. I wanted to
investigate the possibilities that were raised by cryptographers in
the community at the time for differential power analysis techniques
and whether or not one could ascertain bits of the private key stored
in the smart card.

Smart cards have a closed interface meaning you can only access
information on its  "disk" via the API that the smart card provides.
This means you can't just go and pull its disk out and try to read it,
you have to abide by the smart card's rules and talk to it via its OS

(btw if you did try to pull the disk out which is a very tiny thing
that can barely be seen by a microscope you would loose all the data
on the disk before you even had a chance to send a few electrons of
current through it.)




Solving the problem requires the knowledge of the following things:

1.) The energy consumption and time taken for the smart card to do the basic
     mathematical and bitwise operations (+,-,*,/(modulo),|^&,squirt etc...)

2.) The algorithms used in the smart card (modulo-ex (RA), CC etc..)

3.) also if the implementations had any special optimizations that diverged
     from mainstream implementations yet yielded similar results (a possibility
     is adding chains)


1.) to accomplish this step you need a CRO that can sample at about 4
times the operational rate of smart card (shannon says twice the freq), I
found 4 times to be a better proportion, its true you get a lot of
redundancy however they can easily be eliminated through a series of
freq and surge filters.

I found that i not only had to measure energy consumption of the
individual operations but also the operation in group sequences of 2
and 3 operations, meaning all combinations of the above on both integer
and floating point entities.

from this step I began to assemble a table of signatures for
operations which later during real analysis of the smart card going
through its paces I could predict with a guesstimate much better than
50/50 from the energy consumption graph what the smart card was
actually doing at that particular point in time.

2.) this one is a pretty simple one, some smart cards come with their
crypto algorithms pre encoded or pre-built into their circuitry, in
any case the algorithm was known, but the exact implementation was not
known. I never figured out how to blindly (aka black box method)
figure out what the exact implementation was, however my research
allowed for me to program my own smart card or use smart cards with
known programming which made it a bit easier.

knowing what the programming of the algorithm is allows you to build a
flow chart of how the smart card will work when given a particular set
of bits(i.e.: private key), the execution pipeline in an smart card is
very small and simple and does not contain parallel execution units
like microprocessors, this type of architecture allows for a great
deal of DPC to be done.


i.e.: as an example lets say if the LSB was 1 (btw highly unlikely for
composite numbers with factors which are only prime) then the first
operation the smart card would do is an addition(+) or if it was 0 the
first operation it would do would be a subtraction (-), if you measure
the energy consumption for the first clock cycle of operation and did
a best curve fit with the curves calculated in phase 1 you can
ascertain the state of the LSB, continue the analysis and you can
discover bit2, bit3 etc, a caveat to this is that it gets slightly more
difficult and complicated as the algorithm proceeds, but you gotta remember
your objective is not to discover all the bits just some of them...



3.) smart cards are really weak processors of data, they can barely do
512kbit RA and seem to only work in a reasonable amount of time with
CC algorithms. I think thats why many smart card implementations tend
to use CC with smaller key sizes rather than RA with large key sizes
one of the problems is exponentiation of very large numbers, even with
the optimizations of adding chains for modulo exponentiation its still
a very slow intense process for smart cards to undertake. Nevertheless
there are optimizations, so you have to know about them, some of them
are programming based, others are indented for the particular smart
card processor type, I wont go into details of what optimizations I
had to take into account to accomplish phase 2, but needless to say
they were not that difficult to find, if you can find data sheets on
smart card's or talk to marketing people from the smart card
manufactures for info on their products.



These 3 phases allowed me to detect the first 64 bits of the 96-bit
private key stored on the smart card I mentioned above (algo was CC
key exchange and authentication). This is by no means an easy task it
takes a lot of time, and automation of the phases is very difficult,
because of back ground noise, in-consistent energy sources and a whole
array of other ambiguities that are outside the scope of this general
discussion. The lab conditions were near perfect, the real world is
not so accommodating, hence for phase 1 I had to do the calculations
1000s for each group continually changing EM fields around the unit,
tweaking the surge filters etc, until i could statistically say that
the error percentage was low enough for the whole process to produce
acceptable results. Despite all these problems its less-than-brute
force, heaps less! and can be done within hours rather than weeks
months or years...


This type of attack in my opinion should be used in conjunction with
other factoring and DLP algorithms not by itself, in the instance of
factoring because you start discovering the bits from the LSB upwards,
this allows you to construct the private key from one end,hence
further allowing you to guide your sieving algorithms in a much more
efficient manner.


Nowadays some manufactures place "random" operations which really do
nothing other than consume energy and processor time during the
execution of the algorithms to obfuscate the overall energy
consumption of the smart card. Some of these salt like implementations
are really good using dedicated RNG hardware found on the smart card
to decided where, when and what kind of operation to execute
randomly, where as others just use a simple PNG and seed them with the
smart card's processor tick or something easily accessible via the
smart card's interface.


In the lab the setup looks pretty sus, a big HP CRO next to a tiny
smart card-reader, but as the golden rule in cryptography goes "never
underestimate the assets and resolve your opponent may possess to
overcome your security measures". I'm sure someone or some
organization with enough money and time could produce a specialized
sampler that is very small in comparison to the one I was using, build
specialized filters for the noises that are around, have it compensate
for differing energy levels from the power source and also have it
WiFi enabled sending its data back to some receiver where the data is
analyzed, and the key is computed and then copied onto a blank smart
card etc etc...


Anywayz thats been my experience with smart cards and cryptography
analysis with them. Thanks for taking the time to read this,I would
like to end this with a quote from my favorite villain Mr Burns:
"I enjoyed writing this as much as you enjoyed reading it" :D



Regards




Arash Partow

__________________________________________________
Be one who knows what they don't know,
Instead of being one who knows not what they don't know,
Thinking they know everything about all things.
http://www.partow.net










Michael Silk wrote:

> Hello,
> 
>   A bank I am working with provides a Smart Card which is used on our
> clients PC to authorise him to the bank website ... the card stores a
> certificate, and also, i think, a private key.
> 
>   I am trying to discover if, in the case my clients machine is compromised
> (the machine the Smart Card is connected to), what are the possibilities
> of observing the smart card data as it is transferred ?
> 
>   I think that if the smart card reader (scr) and the client program
> (cp) establish a ssl-type connection then the only opportunity to observe
> the private key, certificate, is by inspecting the memory of cp. However
> I don't believe that this (or any?) smart cards to this ...
> 
>   Really appreciate any info anyone has.
> 
> Thanks
>  -- Michael
> 
> 
> 
> Concerned about your privacy? Follow this link to get
> FREE encrypted email: https://www.hushmail.com/?l=2
> 
> Free, ultra-private instant messaging with Hush Messenger
> https://www.hushmail.com/services.php?subloc=messenger&l=434
> 
> Promote security and make money with the Hushmail Affiliate Program: 
> https://www.hushmail.com/about.php?subloc=affiliate&l=427
> 
> 
>