...
> Look. People crack DVDs, HDMI, etc, even when there is little
> monetary incentive. The reason is because there is no way to keep the
> debugger out of a mainstream system -- i.e., they can't hide the
> details, as much as they would want to.
That is, complete information on how it works is available.
I think you are putting the cart before the horse -- hackers *make*
the information available, because hiding information in a binary is
just incentive for some of these people.
> Yet poker sites; where there is a *LOT* of
> monetary incentive remain relative unhacked (I won't claim totally,
> but certainly many have been in long term operation, and are
> profitable and "fair".) How could that be possible if you can't
> generate random numbers?
By not showing the process that generates numbers?
What the -- ?? Obviously you have never thought about security.
Look, the sites have to prevent *INSIDERS* from stealing money through
random number generator disclosure. And security by obscurity is
basically *NOT* security. If the numbers have a discernable pattern,
then shrouding the algorithm or implementation is of no help.
[...] For those sites to
work it is sufficient if the numbers already generated give insufficient
information to predict the next number.
First of all, no its not. Because as much as they would like to keep
their algorithms a secret, disgruntled, unscrupulous or bribably
employees exist.
[...] Whether it is a true random,
a pseudo random or not a random sequence at all. I think a Mersenne
twister with long enough perdiod would be suited perfectly.
MT has a full entropy of about 600 outputs as I recall. That's not
the problem. If the seeding process is deterministic or in other ways
stupid, then you can simply try them out in a brute force manner and
see if the sequence of its outputs match. To make this useful, you
need to fill all the slots of its entropy table (seed) with some
unpredictable start-up pattern, so that all (4 billion)^600 start-up
patterns are possible (with close to even distribution), just to
achieve the full security potential of MT. Even there, because MT has
only had < 10 years of security exposure, you don't know that there
isn't a very simply formula for reverse engineering its state from,
say, 601 of its outputs.
So I think you just don't understand the problem. The peer to peer
approach has the incredible advantage that you can actually, in a very
practical way, cause a reseed to happen well before the 600 outputs,
by forcing the *CLIENTS* to come up with the entropy. The point is
that even if one of the clients could be compromised, or somehow
produces far less entropy than you think, there are plenty of others
plus the server to contend with. So if you do that, then ideally *NO*
approach to attacking its security would be practical short of
installing debuggers on all the clients and the server simultaneously.
[...] Moreover,
I do not think the individual players on such sites even get sufficient
information about the numbers generated so far (in many of those games
distributed cards remain hidden).
That's not the point. Look, just google around for this stuff for a
while and you will see. They take this stuff a lot more seriously
than that.