UC researchers: Take antispam fight to the Web
- 08 August, 2007 08:40
Spammers may have an Achilles Heel.
According to researchers at the University of California, San Diego, antispam fighters could really hurt the spammers bottom line, if they target the Web sites used to host their scams rather than simply trying to block the mail server used to send out unsolicited commercial e-mail.
"If there was more diligence in taking down the Web sites, that would have an effect on overall spam," said Chris Fleizach, a research assistant with U.C. San Diego. "A lot of people using spam to advertise their sites aren't well provisioned," he explained, "so focusing on these bottlenecks would have a deleterious effect on the spam campaign."
Fleizach is part of a research team that recently took a close look at the network infrastructure used by spammers. They concluded that while spammers may use many different servers to send out their e-mail, it's a different story when it comes to hosting the Web page that sells the watches or male enhancement products featured in the spam.
In fact, 94 percent of the time, the scam could be traced back to a single Web server. "Most scams are hosted on a single IP address, providing a potentially convenient single point for network-based interdiction either via IP blacklisting or network filtering," the researcher wrote in a paper to be delivered Thursday at the Usenix Security 2007 conference in Boston.
That's because these Web sites are often run by gray market vendors who need to set up complex e-commerce sites, said Stephen Pao, vice president of product management with antispam vendor Barracuda Networks Inc. "Folks who sell diet pills, folks who sell fake watches, they all need to take credit-card numbers," he said. "And they all need to have central data centers."
The U.C. researchers monitored over 1 million spam messages over a one-week period late last year, tracing them back to more than 36,000 URLs (uniform resource locators). Using a data analysis technique called image shingling, they were able to identify 2,334 distinct spam campaigns, hosted on 7,029 computers.
Previously used by search engines, image shingling can break down a screen shot of a Web page and analyze its graphical elements.
Researchers found that nearly 40 percent of the time the spammer's Web site was also being used by another spam campaign, suggesting that these machines are often rented out to more than one spammer.
Although browser makers like Microsoft and Mozilla have spent a lot of effort beefing up their antiphishing filters this past year, there hasn't been a similar effort to warn people when they're visiting a spam-related site, Fleizach said. "There's no broad community effort to go block sites that are linked to spam, and that may be an important effect that comes out of this [research]."
Enterprise users can block spammer's sites by purchasing Web-filtering products such as those that do "intent analysis" of Web sites, said Barracuda's Pao. But "in general the consumers don't have the tools available for them to do it," he added.