[Christophe_O]

P.P.S. Sorry STEADICAMOP I did not notice your above message: this is an obsolete thread! New version:
Enhanced Image Captcha 2.2b

I came here following a reference by Indie2Industry in the Vbulletin.com thread: Spam bots defeat Recaptcha.

I like this hack!

• So simple, no messing with IP databases, etc.
• I believe much new spam is from 'spamborgs' not bots: humans who do not speak English, acting like trained monkeys. People who speak English generally command a higher salary and their time is too costly for this petty spam. That is why I believe this program will be effective.
• Allows each forum to use different images, thus more difficult to program a robot or train a human to answer correctly at numerous forums.
• Much better for the legitimate visitor than ramping up the obscurity of character recognition.

I am not ready to use this hack yet (my forum just started) but voted anyway as MOTM. I am posting this reply to add a few tips and suggestions, based on my testing today at Steadiforum.com (3.6.2) and TheBlueRoomForum.org (3.7.1).

• Test the 'back' button after clicking the wrong image with Firefox (20% traffic) as well as Explorer (about 80%). Test both the back button on the browser and clicking the word 'back' on the error page. (Some bugs here. Maybe the hack needs automatically to 'refresh'.)
• With Explorer 7 and high speed cable, some of the images at Steadiforum.com did not readily download to my browser. I suggest that all the images files be kept strictly below 4k.
• There are some people from every country who speak English well enough to be good forum members. But may not know Homer Simpson. I suggest keeping the labels generic such as "bald man".
• I don't know if blind people using 'aural readers' can navitate Vbulletin. But if so, they should be directed to a contact form for sign-up.
• I suggest that maybe the directional words such as 'bald man' also be converted to an image? That way, illiterates can not compensate by using translation software. Also 3 directional variations for each image. Example: bald man, hairless person, cartoon man.

P.S. I think this program works well partly because it is not commonly used. This program seems to be especially effective against the current wave of semi-human robots who have trouble reading English.

However, if more widely used, then perhaps added complications may be needed. The choices are limited to about 15, and we must allow 3 tries to be human-friendly. Thus, robots can get through 1/5 the time by guessing. When guessing wrong, they cycle through all other forums, returning the next day until they finally will get through. Maybe this could be overcome by shutting down any IP which sends a guesser three times in three days.

Another option would be to create 2-3 multiple-choice options instead of just one. I.e., the verification page might consist of 2 or 3 challenges to choose the correct image from a row of 7-10 images. The visitor could be asked to check a box beside the correct image, 2-3 times. Still simple for humans, and only requiring a download of 14-21 images, but offering 49-343 possible answers.

P.P.S. Sorry STEADICAMOP I did not notice your above message: this is an obsolete thread! New version:
Enhanced Image Captcha 2.2b