New Method Traps 'Fast' Worms

Researchers at Penn State University have launched a startup to sell their new antivirus and anti-worm technologies. The recently formed Day Zero Systems first plans to sell to antivirus companies the new technology it has developed for identifying and blocking worms.

The Proactive Worm Containment (PWC) approach developed by the researchers is supposed to augment traditional signature-based worm and virus detection, as well as so-called rate-limiting technology. The researchers have applied for a provisional patent for PWC, which uses anomaly detection, not signatures. It looks at packet rate, frequency of connections, and the diversity of connections, and it can find and detain a worm within milliseconds of a cyber attack.

Peng Liu, associate professor of information sciences and technology at Penn State and the lead researcher on the PWC project, acknowledges that anomaly detection isn't new. But the difference with PWC, he says, is it doesn't generate false positives -- it releases legitimate hosts that get temporarily quarantined. "The novelty of PWC is that it can unblock those mistakenly contained hosts very quickly," he says. "Others cannot do this."

Existing rate-limiting technology just slows infected hosts, he notes. And PWC can also find worms that are hiding out in memory because it doesn't just scan the disk, he says. "Our technology detects every packet going out of the network."

But security experts say PWC is just another spin on anomaly detection, which has failed to catch on due to its quirks and resource-intensive nature.

"There are literally hundreds of anomaly models that all look great on paper. But as soon as you deploy them in a real enterprise setting, you find thousands of idiosyncrasies that set the anomaly model off," says Thomas Ptacek, a security researcher with Matasano Security. A proxy server, for instance, would set off any rate and diversity of connection anomaly.

And anomaly detection is not exactly new: Arbor Networks, Lancope, and Mazu have offered this technology for several years, he notes.

Randy Abrams, director of technical education for AV company Eset, says his company today uses heuristics along with signatures. "The shortcoming of the Penn State approach is that a worm can compromise the system before invoking its replication routine," Abrams says. "This means additional backdoors, spyware, rootkits, and other malware can be installed on the computer. It would be when the worm enters a fast replication phase that the Penn State technology kicks in.

"It sounds like a reasonable layer in a defense, but it targets one specific part of the problem."

Few "clever" worm detection and containments schemes see the light of day, notes Matasano's Ptacek. For one thing, enterprises don't want to deploy technology that blocks traffic. "Any time a packet is dropped, the enterprise not only wants to know why it was dropped, but also to have been able to predict [it]."

Secondly, worm detection schemes require manpower, he says. "Every worm detection scheme so far requires at least one full-time person to tune and maintain." And finally, he says, "fast" worms aren't a big priority today. "We haven't seen a serious fast worm outbreak in years, and I think enterprises have more pressing problems right now."

But Liu says most worms are fast worms, statistically. "Also, the damaging worms are all fast worms." The technology could miss slow-spreading worms, but those would probably be caught by signatures or other technologies, says Liu.

The idea is for the PWC to be included as an add-on in AV products or firewalls, for instance, he says.

— Kelly Jackson Higgins, Senior Editor, Dark Reading