New security software developed by researchers at the University of Melbourne will bring to market an appliance that can adapt to changes in network traffic and block security breaches before they happen.

The "smart Internet security" software, developed in conjunction with start-up IntelliGuard IT, will also target cybercriminals and organized crime syndicates that crash Internet business Web sites.

IntelliGuard IT's CEO Philip Mulley said the company's approach to securing Internet services is unlike any security offering available today, because most security products are static - like a firewall or anti-virus software - and need the problem to come to them.

"Static tools recognize bad data, but this technology goes looking for problems that may be in the process of occurring, for example, before a denial-of-service attack there may be a reconnaissance attack," Mulley said. "Our technology will recognize this and also distinguish between good customer-related data and attack data, even during a massive denial-of-service attack."

Mulley, formerly Cisco's general manager of network and information security, said the software is highly intuitive, intelligent, and will make decisions on the "goodness and badness" of data.

Only known as "Intelliguard", the as-yet nameless security appliance is likely to appear at year-end in the form of an appliance for carriers through to the enterprise edge.

"This could even make obsolete network security perimeter products," Mulley said.

Professor Rao Kotagiri, the university's head of the Department of Computer Science and Software Engineering, said it is a critical time for Internet security as more businesses prepare to migrate towards VoIP and video-based systems over the next 12 months.

Kotagiri said the software, now patented in the US, does not use pattern-based protection but looks at profiles of various IP-based sources and can differentiate them.

"It's almost like a pipe where dirty water goes in and clean water comes out - good packets go through and the others are dropped," Kotagiri said. "Most existing products use pattern or statistical approaches. Our methods are statistical but very proprietary and will give better reliability.

The research is based on Dr Tao Peng's PhD project from the university's Department of Electrical and Electronic Engineering in collaboration with Dr Chris Leckie and from Computer Science and Software Engineering.

For more enterprise computing news, visit Computerworld. Story copyright Computerworld, Inc.