Skybuck's Racing Random Number Generator

A random number generator which can only produce one "truely" random number
is pretty useless.

Many "truely" random numbers are needed for preventing attacks from
attackers.

System Time api is limited to 32 bits last time I checked unless one uses
the high performance timers/etc.

The hashing is irrelevant.

It's just a method to transform the original random number.

The attacker's goal would be to predict the original random number and once
he can do that successfully it's only a matter of taking the number through
the hash a couple of times until a match is found.

The matches can be anything.

Suppose america decides to use the random number generator to bomb the
planet... and they don't want anybody to know where the next bomb will fall
then it is important that the numbers are truely random.

An attacker however knows which last locations where bombed.

So this is what I mean with "matches".

The attacker can use the matches to reverse engineer it.

I can take/guess a time... and then take that time through the hash to see
if it produces one or multiple of these locations.

If it does the attacker might be on to something. He might have found one of
the original times etc.

From that moment on it becomes easier for the attacker to "zoom in" on the
range

The rest is history LOL =D

A difference of 1000 milliseconds is not that much

Just 1000 runs necessary to find a matching hash.

Bye,
Skybuck.

In article <13345$4773a77c$541983fa$8374@cache2.tilbu1.nb.home.nl>,
spam.RemoveThis@hotmail.com says...
> A random number generator which can only produce one "truely" random number
> is pretty useless.
>
> Many "truely" random numbers are needed for preventing attacks from
> attackers.

As you should have been able to tell, he means "one at a time". As he
also described, if you want strings of random numbers, just use the
output of one iteration to seed the next one.


>
> System Time api is limited to 32 bits last time I checked unless one uses
> the high performance timers/etc.
>
> The hashing is irrelevant.
>
> It's just a method to transform the original random number.

Yes, but to something that cannot be used to determine the original one.
This is a standard, and mathematically proven, technique.


>
> The attacker's goal would be to predict the original random number and once
> he can do that successfully it's only a matter of taking the number through
> the hash a couple of times until a match is found.
>
> The matches can be anything.

Yes, but for high security uses, instead of using 32 bits and its
resulting 4 Billion possible results, an equivalent algorithm will use
thousands of bits, and the resulting enormous number of possibilities
make a brute-force approach (which is what you are describing)
impractical. There was never any claim that this algorithm was adequate
for truly high-security needs, but they are adequate for normal every-
day random number generators and routine encryption and obfuscation
needs.

.....

--
Remove the ns_ from if replying by e-mail (but keep posts in the
newsgroups if possible).