Slightly Shady SEO | Blackhat is Back, Baby.

There’s about a million different captchas out there, and there’s some way to profit from each and every one of them. The trick is creating re-usable software that can handle as many types as possible. Here I’m going to show some common types of captcha, and the techniques you can use to break them.
Note: These techniques are on how to clean up the captcha, not on how to solve it per se. Solving is just a system of tests you must think up your self, or by training. I’m going to have something more on that in a later post.

Captcha 1: Dots in the Background/Swirls/Random Dashes

• This can be solved by a basic and common technique. You scroll through all of the pixels, checking each one. For each pixel that is colored, you check 1 up, 1 down, 1 to the left, and 1 to the right. If any of those has a color, then you move on to the next pixel. If none of those have color, it’s a rogue pixel, and should be set to white.
• If it’s like the third captcha pictured above, then you should run the procedure 2-3 times to eliminate the swirls (small portions of the circles will be picked up each time.

Captcha 2: Letters in the Background

• Personally, I despise these captchas
• The first step is to go through each pixel, and determine how close it is to white. Find a set amount(Maybe 2 out of the 3 values in the 255,255,255 color scheme should be >200, and the last should be >125) of whiteness that will be eliminated, and converted completely to white.
• Convert everything else to black.
• Run the test detailed for the first type of captcha, to clean up any rogue dots that survived the test.
• When splitting these letters up, make sure to record the start and end points for each letter, as they sometimes overlap on the X axis(though not physically), and this becomes important when determining the orders of the letters.
• Writing the tests for each letter of these is relatively easy. They don’t twist or bend too much, and once they’re converted to black, life is good.

Captcha 3: Fading Gradient Background

• I’ll be honest, these ones are not my expertise, but they’re simple enough.
• The letters on these don’t touch the top, or the bottom of the image, so first, scroll through left to right at the top and bottom obtaining color samples.
• After that, for each horizontal pixel you check, compare the color of each vertical pixel in between the 2 points. If the color is too different, convert to black. If it’s not different, convert to white.

Captcha 4: Strikethroughs

• These look simple enough, but are not for beginners.
• This technique is not perfect, and will alter the letters somewhat.
• Scroll through each column vertically, and mark any black pixels present on a column that has less than 2 black pixels, and connects to less than 10 or so black pixels (horizontally and vertically) within 3 pixels on either side as an arbitrary color(say, red for this example)
• Use the y=mx+b format of a line, and connect all broken red pixels to the nearest pixel on the right with red, effectively recreating the line
• Turn all red pixels white.
• Scroll through one more time. If there is any white pixel that has black pixels on both the top and the bottom within 2 pixels, then set it to black. That should recreate the letters.
• Pray.

So that’s how to clean up captchas for later recognition. Later, I’ll write an entry about how to identify the individual letters.

Posted on Friday, October 26th, 2007 at 10:38 am in the category:captchas, seo. Comment RSS 2.0 feed. Comment it , or trackback this post.