.NET [Showoff] Facebook Captcha OCR

[.NET] [Showoff] Facebook Captcha OCR - Printable Version

+- Hack Forums (https://hackforums.net)
+-- Forum: Tech (https://hackforums.net/forumdisplay.php?fid=444)
+--- Forum: Coding (https://hackforums.net/forumdisplay.php?fid=151)
+---- Forum: Software Development (https://hackforums.net/forumdisplay.php?fid=118)
+---- Thread: [.NET] [Showoff] Facebook Captcha OCR (/showthread.php?tid=4855326)

Pages: Pages 1 2

[Showoff] Facebook Captcha OCR - Niǌa - Jun 7, 2015

So I was working on a Facebook cracker, but a dictionary attack was impossible due to the captchas that show up after 10 attempts. I wrote a function to remove the distortion from the images and then trained Tesseract a bit so it can read the processed images better. The accuracy was around %20, but since Facebook still accepts the captcha response if 1 character is incorrect or missing, the total accuracy is around %40.

Example of an input image:
[Image: Cj1vEQu.jpg]

Obviously Tesseract would not be able to read the text from the image before processing.
Example of a processed image:
[Image: 5Dirska.png]

Tesseract successfully read this image as Xp7bad.

The method I used to process the images is pretty simple. I made a 'selection' of all black colored areas and shrunk them, and then resized them. By shrinking enough all lines are removed, but not the letters as they are bigger. And by enlarging the selections again the letters come back to their original size so they can be read easily.

Combining this with Tor networking so I can get 'unlimited' high anonymity, high speed proxies from the Tor network, I can make one fast, 'proxyless' Facebook cracker. If I can't sell this cracker to anyone for a reasonable price I will probably release the code, including the OCR and image processing code in this section for free.

Also made a captcha OCR library for Amazon. No processing was needed since they are already easily readable. Just a bit of training and the accuracy was around %95. Amazon is proxyless as well. 🙂

* - Abhax - Jun 7, 2015

(Jun 7, 2015 06:48 PM)Tyranitar Wrote: So I was working on a Facebook cracker, but a dictionary attack was impossible due to the captchas that show up after 10 attempts. I wrote a function to remove the distortion from the images and then trained Tesseract a bit so it can read the processed images better. The accuracy was around %20, but since Facebook still accepts the captcha response if 1 character is incorrect or missing, the total accuracy is around %40.

Example of an input image:

Obviously Tesseract would not be able to read the text from the image before processing.
Example of a processed image:

Tesseract successfully read this image as Xp7bad.

The method I used to process the images is I made a 'selection' of all black colored areas and shrunk them, and then resized them. By shrinking enough all lines are removed, and by enlarging again the letters come back to their original size so they can be read easily.

Combining this with Tor networking so I can get 'unlimited' high anonymity, high speed proxies from the Tor network, I can make one fast, 'proxyless' Facebook cracker.

Also made a captcha OCR library for Amazon. No processing was needed since they are already easily readable. Just a bit of training and the accuracy was around %95. Amazon is proxyless as well. 🙂

Awesome! I can't wait to see the outcome of this project!

* - Niǌa - Jun 7, 2015

(Jun 7, 2015 06:52 PM)Abhax Wrote: Awesome! I can't wait to see the outcome of this project!

This is the outcome. All I can do is train it so it is more accurate. Or are you talking about the cracker? If all works well I can (hopefully) bruteforce accounts, as the only limitation I have encountered so far is this bloody captcha.

* - NETExplorer - Jun 7, 2015

I'm surprised a tarantula could make this, let alone an obese one
GGOP

* - Abhax - Jun 7, 2015

I mean the actual cracker lol.

* - Only3 - Jun 7, 2015

(Jun 7, 2015 07:10 PM)PatPositron Wrote: I'm surprised a tarantula could make this, let alone an obese one
GGOP

its tarantalula or u mean taranlalula 😊

* - NETExplorer - Jun 8, 2015

(Jun 7, 2015 08:44 PM)Only3 Wrote: its tarantalula or u mean taranlalula 😊

Pretty sure its tarantula
[Image: q8th.png]

* - Niǌa - Jun 8, 2015

(Jun 8, 2015 08:40 AM)PatPositron Wrote: Pretty sure its tarantula

Damnit, should have switched my Facebook to private. Now everyone can see my photos on Google. :(

* - Nicktron - Jun 8, 2015

(Jun 8, 2015 09:14 AM)Tyranitar Wrote: Damnit, should have switched my Facebook to private. Now everyone can see my photos on Google. :(

At least you have acheived the rank: Family, not forever alone like me. :(

* - Niǌa - Jun 9, 2015

(Jun 8, 2015 10:26 AM)Nicktron Wrote: At least you have acheived the rank: Family, not forever alone like me. :(

Its okay Nick, one day you will achieve this rank.

* - Nicktron - Jun 9, 2015

(Jun 9, 2015 10:51 AM)Tyranitar Wrote: Its okay Nick, one day you will achieve this rank.

:') thanks #hopeforthenope

* - 24/7 - Jun 9, 2015

Impressive work, especially the 95% amazon accuracy 🤝

* - Niǌa - Jun 9, 2015

(Jun 9, 2015 12:22 PM)24/7 Wrote: Impressive work, especially the 95% amazon accuracy 🤝

They are really easy:
[Image: dIDNaIx.png]

* - bluntly - Jun 9, 2015

Acutally a pretty nice piece of coding you have there :D I wouldn't know how to do that because im currently still learning VB.NET :D I have a question...Why do you code? I do it for fun but mostly for having that great feeling that everything actually makes sense and you actually DID that :D

* - Niǌa - Jun 9, 2015

(Jun 9, 2015 01:27 PM)StrangerDanger Wrote: Acutally a pretty nice piece of coding you have there :D I wouldn't know how to do that because im currently still learning VB.NET :D I have a question...Why do you code? I do it for fun but mostly for having that great feeling that everything actually makes sense and you actually DID that :D

I do it for the same reasons. But also one more reason: money. Call me greedy but people pay good money for such a project.