in a report based on four years of Web data, Google researchers found that attackers are deploying highly specialized evasion and obfuscation techniques to stay one step ahead of the defenders.
In "Trends in Circumventing Web-Malware Detection," Google analyzed roughly 160 million pages on 8 million Websites and concluded that the bad guys are adapting to new browser and user behaviors and innovating new techniques. Attackers are getting around malware detection and collection technologies, such as virtual machine client honeypots, browser emulator client honeypots, domain reputation tools and antivirus software.
Socially engineered malware, the type that tricks users into visiting a site or downloading a file, made up less than two persent of all malware observed by Google. While the volume of social engineered malware has been rising over the past few years, Google's security engineers said it's still a small piece of the overall landscape.
Instead, attackers are taking advantage of vulnerabilities in software and Web applications. The lifespan of a given vulnerability is still quite short, Google's security team wrote in the report. Most vulnerabilities are only exploited for a little while until they are no longer useful, and the attackers move on to the next bug.
"Our analysis of which vulnerabilities are actively being exploited over time shows that adversaries quickly switch to new and more reliable exploits to help avoid detection," wrote Google engineers Lucas Ballard and Niels Provos on the Google blog. Ballard and Provos were among the seven authors of the report.
A "prominent" exception was the MDAC vulnerability, which is present in most exploit kits, according to Ballard and Provos.The buffer overflow bug in the Microsoft Data Access components (MDAC) can be exploited by an attacker to take complete control of an affected system. An attacker could then install programs; view, change, or delete data; or create new accounts with full user rights.
Attack sites are increasingly becoming sophisticated, capable of scanning victims' computers to identify malware-detection systems. For those systems, the sites display normal content, and load malicious payloads for unprotected regular users. Other sites identify certain IP addresses as belonging to the systems and shunt the traffic to a benign copy of the page instead of the drive-by-download version.
There were more than 200,000 sites infected by cloaking domains in August 2009, the report found. A large-scale attack known as Gumblar also got past Google's scanners around that time.
"In 2008, we discovered that some malware domains no longer returned malicious payloads to our system but still did so to users," Ballard and Provos wrote, adding the team developed a way to detect IP cloaking. "IP cloaking contributes significantly to the overall number of malicious web sites found by our system," they wrote.