June 08, 2009, 5:04 PM — NEW YORK CITY -- As more employees visit social networking sites while at work, network managers are seeing a rise in accidental malware infections known as drive-by download attacks.
Cybersecurity researchers trying to stop users from inadvertently compromising their machines have come up with a novel idea: Give them PCs running virtual machine software so they can act as sensors that detect malware infections and prevent them from infecting enterprise networks.
The idea was developed by George Mason University's Center for Secure Information Systems (CSIS) in conjunction with Northrup Grumman Information Systems.
This PCs-as-sensors approach was outlined at the Cyber Infrastructure Protection Conference held at the City College of New York last Friday.
Anup Ghosh, Research Professor and Chief Scientist at CSIS, says perimeter security measures such as firewalls and antivirus software fail to catch most drive-by download attacks. He says what works is for users to run their Web browsers on virtual machine (VM) software, which acts as a buffer so that malware is isolated from the host machines.
Ghosh calls turning users into sensors that protect enterprise networks a "game-changing" approach to network security.
"Users get infected by visiting Web 2.0 sites," Ghosh explains. "Trusted Web sites are now compromised. It's about Web 2.0, and it's about sites where users are contributing content. Users can put up Java scripts as easy as HTML. There are lots of infections now coming from Facebook and Blogspot. End users don't need to click on a link and follow a trail. With a drive-by attack, there's no user duping required. You just visit your favorite Web site and get hit by software loaded by someone else."
Ghosh's approach is called Internet Cleanroom, which creates single-use VMs on demand when needed for Web browsing and then deletes after use. Internet Cleanroom is being commercialized through a start-up called Secure Command.
"We're looking at how to take our end users -- who are currently our enemy -- and turn them into sensors," he says. "We're looking at turning every user into a collecting sensor to see what's going on out there. We're using the browser because it's the one piece of software that everyone uses. It's the one piece of software that gets attacked."
Ghosh's approach involves running the browser on a VM that is instrumented to function as a sensor rather than running it natively. The sensors provide information to a database that records malware attacks.
"We see exactly what sites are corrupting that virtual environment," Ghosh explains. "We can look at change detection algorithms that note when a Web site is doing something malicious. This changes the paradigm. Instead of trying to clean up an infected desktop machine, we're turning users into intelligence collection."
Ghosh says the approach requires some overhead since users need to run VM software on their desktops.
Ghosh says Internet Cleanroom is a more promising approach to drive-by download attacks than signature-based approaches used today.
"Every Web site where a user gets infected, we capture the attack. Very rarely is there an existing signature," Ghosh says.
