/sci/ - Science & Math » Thread #11036727

176KiB, 1080x1350, 1569255429247.jpg

View Same Google iqdb SauceNAO

Machine learning on a dataset of leaked passwords

Anonymous Sun 06 Oct 13:41:58 2019 No.11036727 View Reply Original Report

Quoted By: >>11036742 >>11036855 >>11037030 >>11037037 >>11037103 >>11037135

Could that technically generate new password datasets efficiently?
If so that would be an easy way to earn some quick cash

Anonymous

Anonymous Sun 06 Oct 2019 13:51:04 No.11036740 Report

Quoted By: >>11036759 >>11036821

Do you know what machine learning is? It's statistics and algorithms. You would need to find a statistically significant relation between usernames and passwords to do what you propose, and no such relation exists.

Anonymous

Anonymous Sun 06 Oct 2019 13:52:02 No.11036742 Report

Quoted By: >>11036759

>>11036727
You mean it would be a unsupervised dataset? If so, why would you need an AI which generates new passwords? If not, what would the labels be?

Sure some sort of GAN solution where one NN generates passwords, and another needs to determine whether the password was generated or real. I guess that would enable the password generator to become an "educated guesser" which could help in brute forcing, but there are more efficient ways of brute forcing than using a machine learning.

Anonymous

Anonymous Sun 06 Oct 2019 14:00:28 No.11036759 Report

Quoted By: >>11036767

>>11036740
You don't understand my question.
Also it would be plain stupid to find such relation.
>>11036742
I could technically tunnel the outputs to hashcat to datasets with some hashes to crack.
I don't know how efficient that would be

Anonymous

Anonymous Sun 06 Oct 2019 14:07:57 No.11036767 Report

Quoted By: >>11036769

>>11036759
Well, what are you trying to achieve? Crack password encryptions?

Anonymous

View Same Google iqdb SauceNAO 1556576092391.png, 341KiB, 501x624

Anonymous Sun 06 Oct 2019 14:09:45 No.11036769 Report

Quoted By: >>11036779 >>11037182

>>11036767
Generating password lists I could sell
Or maybe some password safety validation program

Anonymous

Anonymous Sun 06 Oct 2019 14:14:34 No.11036779 Report

Quoted By:

>>11036769
>Generating password lists I could sell
A password mocker could be pretty cool, not sure if it's useful for anything though

>Or maybe some password safety validation program
Don't think you'll get much better validation than the $n^m$ safety standard, where n is amount of characters used in password set (a-zA-Z0-9symbols etc) and m is amount of characters used in the password.

Anonymous

Anonymous Sun 06 Oct 2019 14:31:12 No.11036821 Report

Quoted By:

>>11036740
>no such relation exists
Yet to be proven. For all we know there might be a sets of relations.

Anonymous

Anonymous Sun 06 Oct 2019 14:49:57 No.11036855 Report

Quoted By:

>>11036727
Microsoft had a program called telepathword that would guess the next letter in your password it was pretty good.

Anonymous

Anonymous Sun 06 Oct 2019 16:22:18 No.11037030 Report

Quoted By: >>11037133

>>11036727
It all depends in the population. Their variety and set point of origin. To be honest i feel that there's gotta be a big enough database and even so it would inherently be endemic to the area where you got it from

Anonymous

Anonymous Sun 06 Oct 2019 16:24:04 No.11037037 Report

Quoted By:

>>11036727
what the hell is going on with her glasses?

Anonymous

Anonymous Sun 06 Oct 2019 16:48:33 No.11037103 Report

Quoted By:

>>11036727
What? You want to take a dataset of passwords and create new AI generated passwords similar to them? For what reason? How would you make cash out of that?

Anonymous

Anonymous Sun 06 Oct 2019 16:58:38 No.11037133 Report

Quoted By:

>>11037030
Demographical password-analysis. Nice. That would be cool, would enable higher chance brute force attacks against targets. Hard to develop a dataset of personal information and their password though

Anonymous

Anonymous Sun 06 Oct 2019 17:00:07 No.11037135 Report

Quoted By:

>>11036727
Using machine learning to generate passwords is pointless. A typical "poor" password is one with meaning to the user e.g. names of family members, birthplaces etc., which you cannot know unless you have some additional data from which to generate, so you'd have to end up generating random passwords. And when it comes to generating random password you're best of just iterating through a dataset and mutating it etc.

Anonymous

Anonymous Sun 06 Oct 2019 17:27:19 No.11037182 Report

Quoted By:

>>11036769
>Generating password lists I could sell
People buy these lists with the intention of logging into systems, anon

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches