Random function is creating the same numbers in 2 seperate instances of compiled script

[cintaskep]

I am using AutoIt to generate 2 random numbers that act as unique seed numbers for an external program.  Even though I have compiled the script into an exe, running 2 instances of the exe simultaneously somehow means that the random numbers generated are exactly the same.  Can anyone shed some light as to why this is happening?

This is the log output of the 2 instances for each exe, showing: $DataFile, Random($Seed1), Random($Seed2)

 

Instance 1:

F:Garbage_SquidGarbageSquid1-1
0x74719660
0x81142742
F:Garbage_SquidGarbageSquid1-2
0x06222057
0x78780455
F:Garbage_SquidGarbageSquid1-3
0x35456052
0x27343546
F:Garbage_SquidGarbageSquid1-4
0x10657744
0x32951169

 

Instance 2:

G:Garbage_SquidGarbageSquid1-1
0x74719660
0x81142742
G:Garbage_SquidGarbageSquid1-2
0x06222057
0x78780455
G:Garbage_SquidGarbageSquid1-3
0x35456052
0x27343546
G:Garbage_SquidGarbageSquid1-4
0x10657744
0x32951169

etc.

 

I can get around this issue by using a unique identifier and scrambling the random number but I'm so curious as to why the random function is behaving this way!

[Melba23]

cintaskep,

Welcome to the AutoIt forum.

My initial thought is that you might be using the same seed - do you have SRandom in the code somewhere?

But it would certainly help if you posted the code you are using to create the random numbers - see here how to do it.

M23

 

[cintaskep]

Hi Melba, thanks for the welcome

I'm just using the Random function to create the variables that are then passed to an external program to do all sort of fun things.  The variables are set like this:

$Seed1_Random = Random(1, 88888888, 1)
$Seed2_Random = Random(1, 88888888, 1)

This issue is that the same numbers are generated when running the exe twice (at the same time)

[water]

IIRC there has been a discussion some time ago about "how random is random". Unfortunately I can't find it at the moment.

Will post as soon as I'm successfull.

[Melba23]

cintaskep,

i can certainly reproduce your results - very interesting.

I am afraid we will have to wait for jchd or another maths guru to come and explain why this happens - far beyond my very rusty maths skills.

However, adding SRandom(@MSEC) and compiling separate instances gave entirely different sequences as I would have expected - even when the 2 were started quasi-simultaneously (one Running the other).  So that might be a solution for the immediate problem.

M23

[cintaskep]

Thanks Melba,

I've added in a Random(@MSEC) and multiplied that into the Random($Seed) to give a truly random number

I will be checking back to see what the guru's have to say about the earlier issue.

[Melba23]

cintaskep,
 

I've added in a Random(@MSEC) and multiplied that into the Random($Seed) to give a truly random number

I am not sure about that approach - the output from Random is as random as you can get so I cannot see why multiplying it again will make it even more so.  What I meant was more like this:

SRandom(@MSEC)

For $i = 1 To 10

    $Seed1_Random = Random(1, 88888888, 1)
    $Seed2_Random = Random(1, 88888888, 1)

Next

Now you start the Mesenne function using a random seed value (well, 0-999 anyway) and I do not get the same sequence from the 2 instances when run. 

M23

Edited May 17, 2013 by Melba23
Typo

[jchd]

PRNG (pseudo-random number generators) work by returning successive values from a very long to obscenely long sequences of numbers which are close to impossible to predict. The classical issue is the start point that the first run of Rand() will choose: it has to be itself strongly pseudo-random and that is one of the most vulnerable target Eve (the eyedropper) will aim to.

Unlike /dev/random from Unices (e.g. Linux) and AFAIK Windows doesn't have a cryptographically strong random seed primitive (or I don't know it, or it's hidden somewhere).

EDIT: above BS proven wrong, see later post.

Rand() in AutoIt is a very good PRNG and is fairly difficult to break in the general case: it's very hard to predict what Rand() will return next after it has been invoked once.

But --and that is a big butt if you allow me to state it that way-- the problem is, just as you experienced, that Rand() initializes itself to something related to time with high granularity. Using sRand() with a compound of processID and msec and something "hard enough to predict and duplicate" will alleviate the issue, but still be vulnerable to attack by a tech-savvy opponent.

The issue shouldn't surprise anyone: it's obviously hard to produce a reasonably random token from a machine relying on fully deterministic behavior. That's why hardware random sources using trully random physical processes in specific hardware are used in serious cryptographic applications.

I also remind that this point has been pointed out on this forum some time ago.

Edited May 17, 2013 by jchd

[guinness]

What jchd said, I learnt from Yashied to add SRandom(@AutoItPID) to the top of the script.

[jchd]

OK I play guilty Windows has it for a long time even if its early security was not on par with expectations: http://en.wikipedia.org/wiki/CryptGenRandom

That would be a good default for SRandom().

[guinness]

You just reminded me >> #2263

[Bowmore]

Just to put another point of view, having the ability to output the same sequence of pseudo random numbers can be very useful when testing and bug fixing. The AutoIt implementation of rand() and srand() seem to me to behave exactly the same as their C/C++ namesakes.

[jchd]

SRandom($SomeConstant) can always be used for repetitive duplicate tests independant of environment just as in C and friends like you rightly say.