Autoit Math Corrupting Data

[Mechaflash]

I suppose the good questions to ask would be:

• At what line(s) do failure occur?
• If failure occurs in a loop, does it always fail at a certain stage? (i.e. 1st iteration, 5th iteration etc.)
• How big are the numbers getting when the failure occurs?
• Is there a certain operation that it's failing on (addition, concatenation, etc)

feel free to add to this.

[nullschritt]

I suppose the good questions to ask would be:

• At what line(s) do failure occur?

• If failure occurs in a loop, does it always fail at a certain stage? (i.e. 1st iteration, 5th iteration etc.)

• How big are the numbers getting when the failure occurs?

• Is there a certain operation that it's failing on (addition, concatenation, etc)

feel free to add to this.

$string = BinaryToString(_QuickLZ_Decompress($string))

I know what t is, and I have said it in the very first post. It is autoit messing up the math and not returning the right results, the numbers get too big and then can't be returned to their original value because of how autoit doesnt support "big numbers", then it gets junk data, and when it runs that through the decompress function, the script crashes.

Note: It fails randomly, the way it works is it does math operations with random numbers, when two or more numbers are too big in one or more of the matricies, the numbers returned during decryption are all wrong.

This causes the above problem. If you use _arraydisplay on $ar after line 60 in the encrypt function, then on $string after line 108, it will display the numerical values of each character both before and after decrypting, by doing this you can visually see the error.

I also even posted the exact lines of code that cause the problem.

for $i=1 to 5
$enc = _mMul($enc,$keys[$ran[$i+1]])
;then here we run the encrypted text (with the randomized key) through 13 more levels or random encryption
Next

[.....]

for $i=0 to $keysx[0] - 3
$string = _mMul($string, _inv($keys[$keysx[$keysx[0]-$i]]))
Next

Edited August 23, 2012 by nullschritt

[jchd]

nullshritt,

I really don't have enough time to dig into your 450+kb code, nor into your multiple threads with incomplete code to run.

But what I can say is that you should clarify your request:

If you expect comments/advices about your homebrew algorithm, say so (but expose it in understandable form first). You would learn a lot doing so, even from the little I've seen.

If you have a problem with math (seems to be around the dot product), just say so and you'll get advices too.

If you have a problem with a specific operation done by AutoIt code (e.g. dot product of matrices giving unexpected result), please post the simplest standalone reproducer possible and you'll get productive answers as well.

For now, the most useful advice I can give you is to stop dreaming of making your own "secure" algorithm and use one instead of the canned solutions proven effective by the security community, both in terms of speed and strength. Everyone can come up within minutes with his/her own encryption algorithm that he can't find a way to break. That doesn't make said algorithm any more secure than ROT13.

Security and especially encryption is incredibly difficult to get right and can very easily have subtle fatal flaws (in algorithm and/or implementation) and there is a good reason why serious algorithms are deeply mathematically involved and scrutinized for years by worldwide experts. Why not use proven solutions?

[nullschritt]

nullshritt,

I really don't have enough time to dig into your 450+kb code, nor into your multiple threads with incomplete code to run.

But what I can say is that you should clarify your request:

If you expect comments/advices about your homebrew algorithm, say so (but expose it in understandable form first). You would learn a lot doing so, even from the little I've seen.

If you have a problem with math (seems to be around the dot product), just say so and you'll get advices too.

If you have a problem with a specific operation done by AutoIt code (e.g. dot product of matrices giving unexpected result), please post the simplest standalone reproducer possible and you'll get productive answers as well.

For now, the most useful advice I can give you is to stop dreaming of making your own "secure" algorithm and use one instead of the canned solutions proven effective by the security community, both in terms of speed and strength. Everyone can come up within minutes with his/her own encryption algorithm that he can't find a way to break. That doesn't make said algorithm any more secure than ROT13.

Security and especially encryption is incredibly difficult to get right and can very easily have subtle fatal flaws (in algorithm and/or implementation) and there is a good reason why serious algorithms are deeply mathematically involved and scrutinized for years by worldwide experts. Why not use proven solutions?

I am working to create my own algorithm, because the ones included, don't offer anywhere near the same level of encryption(and this is a matter of fact, not opinion).

This algorithm, is more secure than any included with autoit. This algorithm works (most of the time).

And the algorithms included are proven secure? You can't be serious about that.

Although I suppose since no one here knows how to help me, I will just modify my algorithm to work alongside a canned solution, and increase it's security.

Thanks anyways.

[Melba23]

nullschritt,

And then you wonder why no-one offers to help you.....

As you feel that this thread serves no useful purpose, there seems only one thing to do.....<click> { © Jos }

M23

P.S. Do NOT start another one - we obviously know nothing about this subject, so there is no point in asking.