Zero-Day Drive-By Attacks: Accelerating & Expanding

The zero-day attack business is no longer just about money, and patching is no longer the best defense.

Similar to a criminal drive-by, the watering-hole attack redirects unsuspecting victims; the difference is that the redirection (usually via obfuscated JavaScript) is placed on a carefully chosen website where the intended victim will likely browse in the course of their daily employment activities. Indiscriminately exploiting victims is pointless for nation state actors, rather it is a select group of targets that must be compromised in order for the attack to be deemed a success (which for them may entail a better-than-average holiday bonus).

Further, because the intended victims’ computers may be fully patched, nation state actors don’t need a full exploit pack. Instead they can rely on one or two zero-day exploits. (A “zero-day” is security industry jargon for exploit code that targets a previously unknown software vulnerability.) Since government resources are exponentially larger than criminals’, zero-day exploits are purchased from third party brokers or developed internally and used in watering-hole attacks to increase the chances of success.

Two such attacks occurred in May. The first campaign compromised the US Department of Labor’s Site Exposure Matrices (SEM) website -- a very specific watering-hole -- and injected JavaScript code which redirected visitors to Naturally, this website was hosting a zero-day exploit for Internet Explorer (CVE-2013-1347). Following successful exploitation a Remote Access Trojan (RAT) was installed on the victim’s computer.

Subsequent attacks occurred in the same fashion days later when oil and energy company websites were modified to host redirection code. Ten oil/energy sites redirected victims to three different websites hosting exploits. In fact the same Department of Labor Internet Explorer zero day exploit was used in tandem with a Java (CVE-2012-1723) and Firefox/Thunderbird (CVE-2013-1690) exploit. While a zero-day exploit doesn’t remain zero day for long, it is a powerful tool with plenty of potency for quick and targeted campaigns.


Unfortunately the use of zero day-exploits in drive-by attacks appears to be accelerating. In the past two months different zero-day exploits for Internet Explorer were discovered as part of larger strategic web compromise attack campaigns. In the most recent attack a RAT was installed on victim computers and in October Microsoft released a security advisory citing a different Internet Explorer vulnerability that was actively being exploited in Asia.

It’s evident that governments, businesses, and individuals are all at risk for drive-by attacks. When dealing with the criminal set and their exploit packs the answer has always been, patch! Since exploit packs historically bundle large amounts of shell code corresponding to known vulnerabilities, the most efficient method for "p0wnage" prevention was a robust vulnerability identification and security patch management program. Zero-day exploits make this defensive strategy obsolete. So the question becomes what is the answer when comprehensive patching is no longer the solution?

A sensible answer is behavior scoring because there are plenty of common malicious indicators between recent attacks. One practical way to implement the scoring is via a web proxy, specifically to fetch and preview web content before serving it to the requestor. The presence of obfuscated JavaScript code, redirection tags, shell code, and dynamic DNS domains can all be scored, and any content above the tolerance threshold should be rejected before it impacts the end user. Nevertheless, nation state attackers’ behaviors and methodologies will evolve and new defense strategies will need to be implemented.

Finally, it’s not the end of the world if a watering-hole attack succeeds, so long as network (and ideally host) security monitoring programs detect the breach before the company or agency’s intellectual property crown jewels are removed.

Drive-by attacker’s planning and timing can’t be prevented, but we can remove the weapon’s effectiveness.