Millions exposed to malvertising that hid attack code in banner pixels

The malicious script is concealed in the alpha channel that defines the transparency of pixels, making it extremely difficult for even sharp-eyed ad networks to detect. After verifying that the targeted browser isn’t running in a virtual machine or connected to other types of security software often used to detect attacks, the script redirects the browser to a site that hosts three exploits for now-patched Adobe Flash vulnerabilities.

“We can say that even some of the other major exploit kits, like Angler and Neutrino, are outclassed by the Stegano kit in terms of referrals—the websites onto which they managed to get the malicious banners installed,” Eset researchers wrote in a report published Tuesday. “We have observed major domains, including news websites visited by millions of people every day, acting as ‘referrers’ hosting these advertisements. Upon hitting the advertising slot, the browser will display an ordinary-looking banner to the observer. There is, however, a lot more to it than advertising.”

The ads promote applications calling themselves “Browser Defence” and “Broxu” and targeted people who visited the news sites using Internet Explorer browsers. The script concealed in the pixels exploited a now-patched IE vulnerability indexed as CVE-2016-0162 to obtain details about the visitors’ computers. Among other things, the script checked for the presence of packet capture, sandboxing, and virtualization software and a variety of security products. Machines that didn’t exhibit signs of the software and contained a vulnerable version of Flash were then redirected to the exploit site, which would serve one of two families of malware. The Ursnif familyis made up mainly of modules for stealing e-mail credentials, logging keystrokes, taking screenshots and videos, and acting as a backdoor. The Ramnit variety of malware offers most of the same capabilities and mainly targets the banking industry.

The attackers took extra pains to ensure the machines being infected didn’t belong to security-savvy people who might detect what was happening. In addition to a check carried out by the script embedded in the ad, a separate check was carried out by the exploit server before going through with the attack. The Eset report didn’t identify any of the sites that delivered the malicious ads. It did say that the people exposed were concentrated in Canada, the UK, Australia, Spain, and Italy, which are the countries served by the affected ad networks. Earlier versions of the campaign from 2014 and 2015 targeted people in the Netherlands and the Czech Republic. The Flash vulnerabilities exploited included CVE-2015-8641, CVE-2016-1019, and CVE-2016-4117.

Update: To execute the hidden payload, the malicious ads load a heavily modified version of Countly, an open-source package for measuring website traffic. That JavaScript extracts the hidden code out of the image and executes it. Because there’s nothing per se malicious in the JavaScript, ad networks fail to detect what’s happening. Referring to an ad located at hxxps://browser-defence.com/ads/s/index.html?w=160&h=600, Eset researchers described it this way:

The index.html loads countly.min.js and feeds the initial parameters to the script. This countly, however, is not the stock library of the open source mobile & web analytics platform you would download from github. It is a heavily modified and obfuscated version, with some parts deleted and interlaced with custom code. This custom code is responsible for an initial environment check. Information about the environment is reported back to the server as XOR-encrypted parameters of the 1x1gif file, as captured in the image above.

The following information about the environment is sent:
systemLocale^screenResolution^GMT offset^Date^userAgent^pixelRatio

After that, the script will request the advertising banner. The server will reply with either a clean or a malicious version, most likely also depending on the previous environment check.

The script will then attempt to load the banner and read the RGBA structure. If a malicious version of the image was received, it will decode some Javascript and variables from the alpha channel

The steganography is implemented in the following way: Two consecutive alpha values represent the tens and ones of a character code, encoded as a difference from 255 (the full alpha). Moreover, in order to make the change more difficult to spot by naked eye, the difference is minimized using an offset of 32.

Researchers from Eset competitor Malwarebytes have published their own write-up of the campaign, which they are calling AdGholas.