Pictures of homeless cats and dogs could soon be helping users get access to secure websites and services.
The homeless cats proved popular at Techfest
By using pictures that humans can easily distinguish, researchers hope to foil computerised efforts to get at valuable web services that perhaps offer e-mail accounts or sell tickets for popular shows.
The research plan to use pictures was showcased at Microsoft's annual Techfest - a conference that gives the software giant's researchers chance to publicise what they are working on.
Also on show at Techfest were Microsoft's latest efforts to improve the way people search for information that may help the company in its battle with rivals Google and Yahoo.
Hundreds of researchers from labs in China, India, the US, and Cambridge in the UK gathered at the company's headquarters in Redmond, Washington for the event.
Although usually restricted to Microsoft workers, this year Techfest was opened up to customers, government officials, and the media.
One of the most popular research projects at Techfest, called Asirra, involved two cats looking for new homes.
Often when users open a new webmail account, or buy tickets online or try to get access to any kind of secure content they are confronted with a test that tries to ensure only humans can login.
These can take the form of a series of letters and numbers shown in a distorted grid - known as a Captcha - Completely Automated Public Turing test to tell Computers and Humans Apart.
These try to ensure that those logging in are human rather than software in a bid to thwart malicious hackers who use automated scripts to get at valuable information or services.
Humans can spot cats but computers may struggle
For instance, spammers often try to get at web-based e-mail systems to use the free accounts to send out junk mail.
Ironically the distorted characters can be recognised by computers quite easily and often fox humans.
In a bid to overcome this John Doucer came up with another way for people to prove they are human.
"What we want is a challenge that is easy for people, but difficult for software" said Mr Doucer, a researcher at Microsoft's Redmond labs, "and such a challenge is telling a cat from a dog."
The system developed by Mr Doucer and colleagues presents a user with a series of random images of cats and dogs, and asks them to tell which is which.
It uses an index of two million images to make it impossible for attackers to classify all the pictures and break the system. The images used are also easy for humans to distinguish but should be much more difficult for computers because cats and dogs look so similar.
The two homeless cats, called Felix and Cleo, feature in the database of images many of which were taken from a US pet adoption website.
The research project has had one immediate success as at least one of the cats on hand to publicise Asirra has found a home.
As well as showing new ways to get at information, also unveiled at Techfest were novel ways to search data.
One project called "Gazing into Web Search" uses eye tracking to study exactly how people use search engines.
The eyetracker reveals how humans cope with search results.
"The idea here is to understand how people look around the search results page, and how changes to the page will, depending on what the task is, change what they do with the information" says Ed Cutrell, a researcher from Microsoft's Redmond labs.
He has used a commercially available eye tracker, that sits below a computer monitor, and reflects infra-red off the users eyes, without them being aware of it.
Mr Cutrell believes changing the amount of information displayed with search results could change how useful they are.
For instance, he said, the snippets of text that sit beneath the title of a website in a page of search results are useful, but can also be a distraction, if a user is looking for a particular site.
"By using this eye tracking technology we were able to determine that you wind up getting sucked into this extra text, and you lose track of the really relevant information" said Mr Cutrell.
Microsoft researchers also had some ideas about improving the way that people search for pictures. Traditional search engines can only hunt down text, so finding photos usually relies on embedded data, or tags of keywords.
But one idea shown at Techfest uses location recognition.
In this project a photo of a street or building sent to a website is compared to a database of street-side photos, and several likely looking results are returned, the user can then narrow down which one they snapped, to pinpoint their location.
"Suppose you are a traveller - you can take a photo of something you think is specific for a location, and submit it to a server" said Xing Xie from Microsoft Research Asia, based in Beijing
"You can see a map, and more information for that area, for example restaurants, and even the menu at the restaurants."
The service could be run on a camera phone, with the results returned as a webpage formatted for the small screen.