And defy the bots of doom
A GUY HAS devised a fantastic site which shows computers haven't a clue when it comes to kittens. And so neither do bots, either.
Oli Warner devised the KittenAuth
page that avoids stupid computers trying to figure out who is a human being by using a daft algorithm. Oli said the idea came as he was devising a plan to ensure people could file their ideas without being cursed by stupid spambots, which are currently something of a curse round the world wide wibble.
As he said: "If a bot got set loose on ThePCspy.com site it would take weeks to clean it all out."
He said he first thought about asking for email authentication but that's a sign up individuals have to make, so he looked at the ways Turing tests were delivered.
He said: "By far the most popular were CAPTCHA based tests that use distorted texts, obscure fonts, and uncontrasting backgrounds... Whereas we humans just learn to read and that's it, OCR is constantly improving and as soon as there is one 'unbreakable' CAPTCHA design out, there's a bot that smashes through it."
He has included some examples from Digg.com, Yahoo Mail and Hotmail,
here - they are always letters and numbers.
He said: "The human brain is still light years ahead of computers when it comes to quickly recognising objects and animals so making people just decide what was a kitten out of a selection of randomly given photos seemed like a much better idea."
So he devised a scheme of images, some of which were correct and others incorrect. The mathematics behind the probabilities of someone "just fluking" the three by three grid is a problem, he said, but there are workarounds that he's implementing that will help that.
While a "font is always a font", users of KittenAuth don't have to use kittens but can choose what they want - plus this system is great for folk with disabilities. This is clever stuff. We like it. Did we mention CPUs and bumblebees yet?
The INQuirer
Threaded Mode