Analysis On December 1, 2005, two email messages were sent from a computer in Western Australia to members of two different human rights organizations. Each email message carried a Microsoft Word document with a previously unknown exploit that would take control of the targeted person's computer and open up a beachhead into the group's network.
The attack failed, as did a second attempt to infiltrate the same human-rights groups a week later, due in no small part to an overabundance of caution on the part of email security provider MessageLabs, which initially blocked the emails based on the strangeness of the Word attachments. The attacks only targeted a single person at each organization and, after the two attempts, never repeated.
Such targeted Trojan horse attacks are quickly becoming a large concern for corporations, the military and political organizations, said MessageLabs security researcher Alex Shipp. The email security provider intercepted 298 such attacks between May 2005 and May 2006, and the threat of targeted Trojans is only increasing.
"If you haven't noticed these attacks and you are a big company, you have likely already been attacked," Shipp told attendees at the Virus Bulletin 2006 conference. "Your problem is no longer how do I avoid being attacked, but how do I find where I've been compromised."
Targeted Trojan horse attacks are quickly becoming a major issue for the antivirus and computer-security industries. Last year, computer emergency response groups in the UK, Canada and Australia warned of such attacks. While the United States Computer Emergency Readiness Team (US-CERT) did not issue a warning, security firms confirmed at the time that US government agencies and companies had already been targeted by such malicious software.
A major problem for large companies, government agencies and other potential targets is that antivirus software is not good at stopping low-volume attacks aimed at single companies. Traditional antivirus programs detect widespread attacks based on matching to a known pattern and do not fare well against low-volume Trojans. And even when they do detect such attacks, the larger volume threats are inevitably moved to the top of the firms' to-do lists, because they affect a larger number of customers, said antivirus industry insiders.
"There is no value whatsoever in having signature-based antivirus when facing a targeted attack," said Joshua Corman, host protection architect for Internet Security Systems (ISS). "We, the AV industry, haven't turned the corner in being able to detect these attacks consistently."
If a company misses the initial attack, the results can be costly, Corman said. He pointed to an example of one company, a pharmaceutical firm, that got infiltrated by a targeted Trojan attack. The company only realized it had been compromised when some valuable data was encrypted and the key held for ransom. The company had to pay and, after the incident, spent a month cleaning the compromised systems from its offices in three countries. Corman would not name the company, which became ISS client after the incident.
While MessageLabs might detect tens of thousands of copies of a typical mass-mailing computer viruses in a single day, the company is finding, at most, ten targeted Trojans a week, Shipp said. According to the data collected by MessageLabs, more than half of the 298 attacks detected over 12 months consist of a single email sent to a single person at a company. In total, 1,344 emails were sent during the period studied by Shipp. Military agencies, human rights organizations and pharmaceutical companies are some of the types of groups that are being targeted by specifically aimed attacks.
The attacks are also very well researched, Shipp said. One targeted Trojan was sent to five employees at one company - every single person was a member of the firm's research and development team.
"The bad guys have done their homework," Shipp said.
During the 12 months studied by Shipp, the majority of the Trojan horse programs, almost 70 per cent, used a malicious Word document as the vehicle for the attack. That's already changing, with PowerPoint and Excel documents now becoming popular, he said. The one type of document that oddly is not being used by attackers is the PDF format of Adobe Acrobat.
Shipp added that most companies cannot just block the problematic attachments, even if they realize the threat.
"In many cases, even if the company is vulnerable, .doc files are their lifeblood, so they can't block Word documents," Shipp said.
Most of the attacks come from the Pacific Rim, emanating from Internet addresses in mainland China, Hong Kong, Australia and Malaysia. However, one IP address that consistently attacks military installations comes from a computer in California. Shipp believes that the computer could have been compromised as part of a botnet.
In fact, Shipp believes that three major groups are involved. The first and largest group is the most active and uses a variety of different tactics, but most commonly uses a zero-day exploit in the attack. The researcher believes it is also possible that this group could be several independent actors. He feels more confident about the other two groups: One targets only Hong Kong companies with an attack once every three weeks or so, and the other attacks military sites from a computer in California about every two weeks.
The attack data underscores an overall trend in threats. While some hobbyist virus writers undoubtedly still exist, most malware is now written for profit.
"No one other than the kids want to infect a million people anymore," said Graham Cluley, senior technology consultant for antivirus firm Sophos. "You would rather deal with 50 or 100 systems at a time."
Sophos and other security companies are adopting better versions of behavioral blocking software to combat the threat. Traditional behavioral blocking stops a program that attempts a specific action, a technique that frequently flags legitimate programs as potential threats. Sophos instead characterizes programs by a collection of actions attempted by the program in a virtual sandbox and blocks the executable if the actions seems malicious. Called "behavioral genotype protection" by Sophos, the technique has already caught a number of targeted and low-volume attacks, Cluley said.
However, the antivirus industry is still moving too slowly, ISS's Corman said. The Trojan horse sold to private investigators by an Israeli couple took 18 months to detect.
"People in the industry keep talking about the Israeli Trojan horse, because that is one of the few public examples," Corman said. "But that's just one of hundreds, if not thousands, of successful attacks."
This article originally appeared in Security Focus.
Copyright © 2006, SecurityFocus