Simple tests designed to distinguish computers from humans are increasingly being used to clamp down on unsolicited, or 鈥渟pam鈥�, email advertising.
The tests involve deciphering a random word that has been distorted or partly obscured. Though this is a relatively simple task for a person, it remains beyond most computers. It is known as a 鈥渃ompletely automated public Turing tests to tell computers and humans apart鈥� or visual 鈥淐APTCHA鈥�.
A computer will find it hard read the word 鈥榮mile鈥� in this test
Advertisement
In 1950, the British mathematician Alan Turing suggested that if a computer鈥檚 performance was indistinguishable from that of a human, it could be deemed intelligent. But since then researchers have struggled to come up with any machine capable of passing such a Turing test.
The senders of spam email use robot programs 鈥� or 鈥渂ots鈥� 鈥� to harvest addresses from web pages, sign up for free email accounts and spew out millions of unwanted messages each day. So sorting the bots from real people is very useful in tackling this growing nuisance.
One company has developed a system that uses a CAPTCHA test to filter out spam email before it reaches a user鈥檚 inbox. Spam Arrest鈥檚 software requires a sender to complete a simple visual test before their message can be delivered to its intended recipient.
Once the sender has passed the test, all future email messages will be passed on by the program. Programs that automatically send out spam should be confounded by the check.
Sign up
Similar tests are already used by Yahoo! to prevent bots from signing up for their free email accounts in order to send out spam. Hotmail, the most popular free email service, introduced such tests in December.
This does not however, prevent spammers from harvesting email addresses in the first place and sending advertising from other accounts.
One team of researchers, at the University of California in Berkeley, has also devoted its expertise to beating visual CAPTHCA tests. The group has adapted methods for recognising more complex shapes, like human handwriting and human figures, to tackle the test used by Yahoo!.
On 10 December they announced that their software could beat the test 83 per cent of the time. But Yahoo! has since made its test more difficult by increasing the number of characters that must be recognised.