February 04, 2003
The West Australian
By Fran Spencer

AUSTRALIA'S corporate watchdog has launched a $1 million-plus research project aimed at harnessing the power of language technology to crack down on the growing number of internet scams and dodgy websites.

The Australian Securities & Investments Commission (ASIC) said yesterday it had joined forces with the Capital Markets Cooperative Research Centre (CMCRC), the University of Sydney and Macquarie University to develop an automatic internet document classification system known as Scamseek.

The project, the biggest of its kind ever initiated in Australia, will have a budget of "over $1 million" and focus on using innovative language technology to identify possible breaches of the Corporations Act on websites.

According to the project partners, Scamseek aims to determine the potential risk of a site by scanning against public and private databases and assess the risks associated with information on the site.

It will also identify people and companies mentioned on the site and flag sites identified as "above the acceptable risk threshold" for further analysis.

ASIC has been using the web to sniff out scams for some time, participating in "Surf Days" run by the International Organisation of Securities Commissions as well as running its own internet surveillance campaigns.

But ASIC director of electronic enforcement Keith Inman said the internet represented enormous opportunities for scams, putting pressure on regulators to develop "an eye that never sleeps" to seek out disreputable sites.

"For a couple of years now ASIC has sought to be proactive in identifying suspect websites that offer people illegal investments or unlicensed investment advice," he said.

"For us the internet is in some ways a problem - all of a sudden people have a cheap and easy method of reaching a lot of people quickly, but on the other hand it also makes them more susceptible to detection.

"They have to put an offer on the web to entice sufficient numbers of people, and during that period they're susceptible to detection by the regulator . . . it's that weakness we want to take advantage of."

Mr Inman said ASIC had been actively monitoring the internet for suspicious sites with the aim of shutting them down before they ensnared too many victims.

"The downside to that is it's still, despite the availability of good search engines, manually intensive for us because when you search by a set of keywords for your two or three scam websites you will also bring back 1000 legitimate sites," he said.

"So there was the need to automate the process."

ASIC had taken the concept further with the development of an in-house system to identify suspect sites, but recognised the need to partner with industry expertise.

"We'll really know we've been successful when we reduce those 1000 possible sites to 100," Mr Inman said.

Scamseek will use the most up-to-date findings on document classification and new analytical methods for identifying the meaning of words to finger suspect sites.

CMCRC and University of Sydney team leader Jon Patrick said scams run on the internet tended to use "certain words in certain ways with certain characteristics", but could also be cleverly disguised in an attempt to avoid detection.

"(With the linguistics department at Macquarie University) we're using new theories on textual meanings to unravel the deep linguistic features that will enable us to detect scam proposals no matter what surface form of language they use," he said.

The project will also apply a specialist web spider to hunt out potential sites using technology developed by Security Markets Automated Research Training and Surveillance, one of the CMCRC industry members.

Professor Patrick said countries, including the US, the Netherlands, Canada and the UK, had already expressed interest in the project, the first stage of which is expected to take six months.

While the ultimate aim would be to develop a commercial product, Mr Inman said ASIC would be unlikely to see any commercial return from a sale of Scamseeker.