visualCaptcha bot

Automate visualCaptcha, "the best captcha alternative".

Overview

Description

I just wanted to show how easy it is to automate the visualCaptcha solution.

After adding the bookmarklet, go to http://demo.visualcaptcha.net/ and launch it (or just copy the script in the JavaScript console). The right image should be automatically selected.

You can renew the challenge and test it again and again. It will work with a 100% success rate unless new images were added since I computed the images list.

Installation

Add this bookmarklet and run it on a page containing a visualCaptcha (or just paste it in the JavaScript console).

Note: if you use NoScript, you'll have to (temporarily) allow www.myersdaily.org and raw.githubusercontent.com.

(function(){function a(a,b){var c=document.createElement("script");c.src=a,c.onload=b,document.head.appendChild(c)}function b(b){"md5"in window?b():a("http://www.myersdaily.org/joseph/javascript/md5.js",b)}function c(b){"visualCaptchaBot"in window?b():a("https://raw.githubusercontent.com/valeriangalliat/visualcaptcha-bot/master/bot.js",b)}function d(){visualCaptchaBot.main()}b(c.bind(this,d))}());

Actual bookmarklet source:

(function () {
    function addScript(src, cb) {
        var script = document.createElement('script');
        script.src = src;
        script.onload = cb;
        document.head.appendChild(script);
    }

    function a(cb) {
        if (!('md5' in window)) {
            addScript('http://www.myersdaily.org/joseph/javascript/md5.js', cb);
        } else {
            cb();
        }
    }

    function b(cb) {
        if (!('visualCaptchaBot' in window)) {
            addScript('https://raw.githubusercontent.com/valeriangalliat/visualcaptcha-bot/master/bot.js', cb);
        } else {
            cb();
        }
    }

    function c() {
        visualCaptchaBot.main();
    }

    a(b.bind(this, c));
}());

Process

To begin, the list of challenge images associated to their MD5 sum is retrieved with the following shell script:

git clone https://github.com/emotionLoop/visualCaptcha-packagist.git
cd visualCaptcha-packagist/src/visualCaptcha/assets/images

find . -not -name '*@2x.png' -name '*.png' \
    | xargs md5sum \
    | awk '
        BEGIN {
            print "{"
        }

        {
            print " \"" $1 "\": \"" substr($2, 3, length($2) - 6) "\","
        }

        END {
            print "}"
        }
    '

The bookmarklet loads a JavaScript MD5 function, needed to compare an image data to the previously computed list.

Then, the bot.js script will loop through all .visualCaptcha elements and try to fill them.

The target MD5 sum is the precomputed sum for the current challenge; if the challenge says to find a clip, we can lookup for the clip key in the static list.

Then, the steps are the following:

• get each image URL,
• get the binary image data in a blob,
• compute the MD5 sum,
• compare it to the target MD5 sum.

When the sum match, we can just simulate a click on the right image.