Is my CAPTCHA Img generation good enough for AI bots and OCR to not read it
This is an example captcha that was generated by my code i wrote for a university project, i wanted your lots opinion on if i should add other secuirty layers to it to make it harder for AI's or OCR tech to read it the code is at https://github.com/UmbrellaCrow612/CAPTCHA
GitHub
GitHub - UmbrellaCrow612/CAPTCHA
Contribute to UmbrellaCrow612/CAPTCHA development by creating an account on GitHub.
8 Replies
If the bots were written in 2004, then yes. Otherwise .. Probably not
I'm sure the image processor can filter out the pixels based on the brightness
https://doom-captcha.vercel.app/ is the way to go
What could I add to make it harder for bots ?
Is there services to test against
Not that I know of
Maybe make a python script that automates it or write some playwright / selenium if it's web-based
I could train a AI model to try and solve it but would people online do the same ?
Or is there a service they would use that is pretained on text based CAPTCHA
I'm sure there's AI models out there for captchas
Could try poisoning the final image with Nightshade
And/or Glaze
Bots are generally better at people solving static captchas at this point. Task based plus unobtrusive analysis seem to work better but anything that stops a bot is likely going to also frustrate and alienate a human as well.