University researchers gained control of a network of hacked PCs, harvesting 70GB of compromised data from the machines.
Researchers from the University of California exploited a botnet, known as Torpig or Sinowal, and monitored more than 180,000 hacked computers by exploiting a weakness within the command-and-control network used by the hackers to control the computers. It only worked for 10 days, however, until the hackers updated the command-and-control instructions, according to the researchers' 13-page paper.
Still, that was enough of a window to see the data-collecting power of Torpig/Sinowal. In that short time, about 70G bytes of data were collected from hacked computers.
The researchers stored the data and are working with law enforcement agencies such as the FBI, ISPs and the US Department of Defense to notify victims. ISPs also have shut down some webtes that were used to supply new commands to the hacked machines, they wrote.
Torpig/Sinowal can pilfer user names and passwords from e-mail clients such as Outlook, Thunderbird and Eudora while also collecting e-mail addresses in those programs for use by spammers. It can also collect passwords from Web browsers.
Torpig/Sinowal can infect a PC if a computer visits a malicious website that is designed to test whether the computer has unpatched software, a technique known as a drive-by download attack. If the computer is vulnerable, a low-level piece of malicious software called a rootkit is slipped deep into the system.
The researchers found out that Torpig/Sinowal ends up on a system after it is first infected by Mebroot, a rootkit that appeared around December 2007.
Mebroot infects a computer's Master Boot Record (MBR), the first code a computer looks for when booting the operating system after the BIOS runs. Mebroot is powerful since any data that leaves the computer can be intercepted. Mebroot can also download other code to the computer.
Torpig/Sinowal is customised to grab data when a person visits certain online banking and other sites. It has been coded to respond to more than 300 websites, with the top targeted ones being PayPal, Poste Italiane, Capital One, E-Trade and Chase bank, the paper said.
If a person goes to a banking website, a falsified form is delivered that appears to be part of the legitimate site, but asks for a range of data a bank would not normally request, such as a PIN or a credit card number.
Websites using SSL (Secure Sockets Layer) encryption are not safe if used by a PC with Torpig/Sinowal, since the malicious software will grab information before it is encrypted, the researchers wrote.
Hackers typically sell passwords and banking information on underground forums to other criminals, who try to covert the data into cash. While it's difficult to precisely estimate the value of the information collected over the 10 days, it could be worth between US$83,000 to $8.3 million, the research paper said.
There are ways to disrupt botnets such as Torpig/Sinowal. The botnet code includes an algorithm that generates domain names that the malware calls on for new instructions.
Security engineers have often been able to figure out those algorithms to predict which domains the malware will call on, and preregister those domains to disrupt the botnet. It is an expensive process, however. The Conficker worm, for example, can generate up to 50,000 domain names a day.
Registrars, companies that sell domain name registrations, should take a greater role in cooperating with the security community, the researchers wrote. But registrars have their own issues.
"With few exceptions, they often lack the resources, incentives or culture to deal with security issues associated with their roles," the paper said.