Page 1 of 1

spambot capabilities.

Posted: Fri Jan 19, 2007 4:08 pm
by jyhm
Does anyone know the current capabilities of a modern day spambot?
I have googled a bit like this guy Neil Gunton's articles about how
he researched spambots and captcha.net. From what I have read,
spambots have an inability to detect distorted text in an image.

My question is how distorted does it have to be because I am developing
a image generator in PHP for use in an email form but I am designing it
for systems that do not have a freetype library installed and just use the
standard gd type library. See below at the image generated, and is this
distorted enough? PHP image filters do not seem to have a distortion
capability.random image

Posted: Fri Jan 19, 2007 8:09 pm
by Buddha443556
Run it through an OCR program and see if can read it. That would be the technology your trying to defeat.

Posted: Sat Jan 20, 2007 3:56 am
by m3mn0n
Add some random lines that maybe even go through letters and such. Most of the OCR programs that read these images rely on lazy programmers not building strong imagine manipulation into the application behind the image generation.

For example, a popular CAPTCHA system is basically numbers as easy to read as the letters in this post on a solid color background. Someone went ahead and build an ugly OCR just for this system which detects the 3 numbers, then does pixel by pixel scanning (based on colour) until it can guess at the best match.

I tried that OCR out and it read the text perfectly for an image I saved from a blog comment page, but on a tougher image it tanked miserably. So the trick is to distort the text as much as possible, add lines... dots.. colours... blur.. anything that'll make OCR harder, and then hope it's good enough. :)