Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Why Captchas have gotten so difficult

theverge.com

Post reply on HN
Print view View original

161–170 of 218 posts

Re: Why Captchas have gotten so difficult

#161

Post by hombre_fatal »

syrrim wrote: post #157
hombre_fatal wrote: post #140

Earlier quoted context omitted.

> We hate them for the same reason we hate airport security theater; they do not work Well, they absolutely do work. They work so well that they've reduced bot actions by almost 100% on our sites. We wouldn't use Recaptcha if there wasn't abuse on the internet. But, unfortunately, there is. There is a sobering amount of it. I'm actually curious about all these posts suggesting that websites use Recaptcha for no real…

The false negative rate may be very low, this doesn't speak to the false postive rate. Considering the goal was supposed to be "telling humans and computers apart" both kinds of errors are important to consider. Google predicates their captcha primarily on the logged-in user's activity around the web, failing that they use the ip. A human being not logged in and using an open proxy is likely to get shut out completel…

Sure, but that should be a chilling reminder of how bad abuse is on the internet.

Obviously it comes with downsides, but it's a trade-off. Nobody uses Recaptcha for fun.

As I reminded a sibling comment, even HN uses Recaptcha on its login/register page. There's no telling how many fewer spambots we have to deal with every day because of it, yet we're somehow here discussing whether Recaptcha servers a purpose while profiting from it. :)

Re: Why Captchas have gotten so difficult

#162

Post by computerex »

outime wrote: post #24

Google reCAPTCHA is the absolute worst. It makes me solve several puzzles very often, usually when I use a mobile network and I’m not logged in with any Google account. It’s so frustrating that most of the times I find a reCAPTCHA I give up before trying and just go elsewhere e.g. when a site uses reCAPTCHA for sign up or after the first failed login, I’ll most likely skip if I don’t absolutely need to access such we…

It's using you as human labor to label data. That's why you have to solve multiple. It's annoying but greatly helps accelerate research and development since data is king.

Re: Why Captchas have gotten so difficult

#163

Post by computerex »

umvi wrote: post #155

I'm not convinced OCR is as good as humans. I recent did a project making an unauthorized copy of a rare ($2000) book from a university library. I scanned in every page, but tesseract OCR really struggled with pages that started off straight but curved off. I tried lots of preprocessing techniques with limited success. My options were to type it in by hand or rescan that page so the lines were straight.

What a weird way to compare OCR to humans. If a human can't see the writing because the page surface is curving away, they'd adjust the page surface. Likewise, getting decent scans is the most cost and effort effective way of getting good performance in OCR.

I personally found tesseract to be incredibly good, and have even used it in non-traditional OCR applications for doing things like reading signs.

Re: Why Captchas have gotten so difficult

#164

Post by corey_moncure »

computerex wrote: post #162
outime wrote: post #24

Google reCAPTCHA is the absolute worst. It makes me solve several puzzles very often, usually when I use a mobile network and I’m not logged in with any Google account. It’s so frustrating that most of the times I find a reCAPTCHA I give up before trying and just go elsewhere e.g. when a site uses reCAPTCHA for sign up or after the first failed login, I’ll most likely skip if I don’t absolutely need to access such we…

It's using you as human labor to label data. That's why you have to solve multiple. It's annoying but greatly helps accelerate research and development since data is king.

If this is so, how does it know when I've mis-labeled?

Re: Why Captchas have gotten so difficult

#165

Post by tigershark »

outime wrote: post #24

Google reCAPTCHA is the absolute worst. It makes me solve several puzzles very often, usually when I use a mobile network and I’m not logged in with any Google account. It’s so frustrating that most of the times I find a reCAPTCHA I give up before trying and just go elsewhere e.g. when a site uses reCAPTCHA for sign up or after the first failed login, I’ll most likely skip if I don’t absolutely need to access such we…

I am starting to think that I may be a machine given how many time I fail that stupid test... The traffic light is absolutely the worse. Should I select the box that includes the vertical yellow pole? And what about the horizontal one between two opposite traffic lights? And if just a small fraction of the border of the traffic light is in a box what should I do? I can’t wait for a captcha solver as a service honestly...

Re: Why Captchas have gotten so difficult

#166

Post by peteretep »

> The latest version, reCaptcha v3, announced late last year, uses “adaptive risk analysis” to score traffic according to how suspicious it seems; website owners can then choose to present sketchy users with a challenge, like a password request or two-factor authentication

eg: if you're not browsing the web signed into a Google account and allowing all their tracking. Fuck that.

Re: Why Captchas have gotten so difficult

#167

Post by captainbland »

corey_moncure wrote: post #164
computerex wrote: post #162

Earlier quoted context omitted.

It's using you as human labor to label data. That's why you have to solve multiple. It's annoying but greatly helps accelerate research and development since data is king.

If this is so, how does it know when I've mis-labeled?

I think it works by labelling something like half the images, but doesn't tell you which is which. At least that's how the older text recaptcha stuff worked - you could easily identify the pre-labelled text by the kind of distortions they used and just enter any old gibberish for the rest. So actually the training stuff is extra work beyond just classifying you as human

Re: Why Captchas have gotten so difficult

#168

Post by henriquemaia »

ufo wrote: post #160
outime wrote: post #24

Google reCAPTCHA is the absolute worst. It makes me solve several puzzles very often, usually when I use a mobile network and I’m not logged in with any Google account. It’s so frustrating that most of the times I find a reCAPTCHA I give up before trying and just go elsewhere e.g. when a site uses reCAPTCHA for sign up or after the first failed login, I’ll most likely skip if I don’t absolutely need to access such we…

The most annoying thing for me is that I often encounter them in various websites for the Brazilian government, when I want to get somd official documents.

This is how things are done at the Federal University I'm currently attending. I have filed a complaint against the university on the grounds that they're using Federal resources (the students' labor, whose place there is being paid by the government) to enrich a private, international corporation (Google).

This is wrong on so many levels.

There's no decision yet, and the process is still running.

Re: Why Captchas have gotten so difficult

#169

Post by umvi »

computerex wrote: post #163
umvi wrote: post #155

I'm not convinced OCR is as good as humans. I recent did a project making an unauthorized copy of a rare ($2000) book from a university library. I scanned in every page, but tesseract OCR really struggled with pages that started off straight but curved off. I tried lots of preprocessing techniques with limited success. My options were to type it in by hand or rescan that page so the lines were straight.

What a weird way to compare OCR to humans. If a human can't see the writing because the page surface is curving away, they'd adjust the page surface. Likewise, getting decent scans is the most cost and effort effective way of getting good performance in OCR. I personally found tesseract to be incredibly good, and have even used it in non-traditional OCR applications for doing things like reading signs.

I'm saying that I can read skewed/bending pages easily, but I'm having a super hard time getting tesseract to play nice with such images. I almost always need to rescan.

Tesseract is incredibly good... as long as your lines of words are straight.

Re: Why Captchas have gotten so difficult

#170

Post by sm4rk0 »

chocolatkey wrote: post #139
pmoriarty wrote: post #84

Earlier quoted context omitted.

There's other (mysterious to me) stuff that sites call to Google for, apart from captchas and analytics. As a uMatrix (and former NoScript) user, I've long noticed that many sites make calls to ajax.googleapis.com for I have no idea what. Quite often the site will refuse to work without that. To be fair, a lot of sites make use of javascript from a lot of other sites as well: cloudfront and amazon are common.

As a uMatrix user as well, I have manually whitelisted ajax.googleapis.com. It is actually not tracking scripts, but hosted libs like jQuery, which many many sites use and will not function without. See here: https://developers.google.com/speed/libraries/

There's solution for that: "Decentraleyes - Local CDN Emulation" https://decentraleyes.org/
Post reply on HN