Hacker News Reader

Comment by dlor | Hacker News Reader

I know folks hate the centralization of identity management to the big identity providers, but as AI gets better and better at defeating captcha it's going to get harder and harder for the small, independent ones to operate reliably and securely.

View on HN

"Never use your real name on that internet thing" will soon be a quaint throwback.

We're moving closer to "when you post on the internet, it is signed and big brother knows you did it. Watch your words"

We are already there since many years ago.

I wrote a lot of the systems to break captchas. Currently the "small" (really medium), independent ones are doing way way way better than the big identity providers. Google reCAPTCHA was particularly easy for us to get past.

Shameless plug: http://resolved.dominuslabs.com feel free to reach out for custom work - specializing in accessing gated sites for new search engines to do better web crawling, scraping public data, building better API integrations on top of existing services, etc.

Open Source captcha libraries are non-viable, you can train a NN on a few hours with a simple CNN architecture, there are some papers that describe how to do it.

We didn't have a lot of luck with simple CNN's on modern CAPTCHA's. But more advanced techniques work great on pretty much all of them.

Reply on news.ycombinator.com