chattersec / open-captcha Goto Github PK

I dont know many captchas that deploy this but there is a technique to manipulate ML by using "adversarial image" this can be done quite simply in python here is a pytorch article

As mentioned in pr #1 the /four.js library has over 300 files that we don't even use, removing all this and also converting what is needed to ts is required.

When running code with open-captcha for more than 20 seconds you get the error X connection to :0 broken (explicit kill or server shutdown). im assuming "x conection" refers to the x11 client-server-based graphics system since the majority of the operaions made by open-captcha is graphical.

A shower thought was to generate a canvas object on the server side and then render it on the clients browser this would elimanate server side rendering issues however will require the client to have canvas and html5 active

Currently, it takes 3 seconds to generate and render 10 images. In terms of request time, this needs to be a little faster. Trying to reduce it to 1 second would be a significant optimization, as this is only the first step.

Render time 1: 212.674ms
Render time 0: 844.614ms
Render time 3: 460.767ms
Render time 2: 584.058ms
Render time 4: 719.112ms
Render time 8: 827.102ms
Render time 6: 946.664ms
Render time 7: 1.062s
Render time 5: 1.184s
Render time 9: 1.290s