Can someone explain how OCR works?
 

Don't tell me you type in the words in the box for visual verification I know that much lol.

How is it created? => how does the img match what the server is looking for?

Could it be automated ever?

a lot can automate the simple ocr, but the truely hardcore ocr can only be beaten by a few people, a very few people, and the rest of the hard stuff is done by hand

It's created much the same way automated fan signs are created. Some servers create the img in realtime (once it's requested) while others go from a database of prerendered images. Either way the server knows what info the image is displaying and thats what it looks for in return.

http://sentinel.securibox.net/Caecus.php

http://serv4u.sytes.net/~tonyvegas/ (ocr at the bottom)...

Here's how one of them works:

Original form is requested. A unique key is created and stored against the user (cookie, session, etc). That key contains the text written to the image. When the form is submitted, the text the user entered from the image is encrypted the same way the original key was encrypted, and then compared against what was originally stored against the user.

Not saying that's how they all work, but at least one of them I've seen.

You're probably confusing some terminology, OCR stands for optical character recognition. It's what computer does to recognize an image, not typing words to match generated text. If you want to know about generating text, it's very easy, a simple script just matches up your session to random distorted word.

If you're talking about cracking a protection using OCR, most programs use either neural network or a probability model, both have to be trained for maximum recognition of the protection that you're trying to bypass.

outsource your ocr work fool

www.iwebmasters.com

Good examples so far =)

Depending on how you do it, it will take too long and take too many server resources. If you need cheap img verif work as in FRACTIONS of PENNIES per entry, let me know. NOTE: your mileage will vary based on your app, connection speed, network, etc. But it should be way cheaper and faster than OCR. Contact info is in my sig.