Google’s Bouncer malware detection system might not be as strong as the Chocolate Factory hopes, with a pair of security researchers demonstrating flaws in the system.
Duo Security’s Jon Oberheide and Charlie Miller, preparing a presentation for this week’s SummerCon in Brooklyn, have demonstrated that it’s possible to slip a malicious app past Bouncer’s protection and into the Android market.
The demonstration in video below shows Oberheide accessing the remote shell to, as he puts it, “look for interesting attributes of the Bouncer environment such as the version of the kernel it’s running, the contents of the filesystem, or information about some of the devices emulated by the Bouncer environment.”
The point, Oberheide says, is that by fingerprinting the Bouncer environment, it’s also possible to identify the characteristics that Bouncer is looking for to identify malicious apps – and to avoid triggering its alarms. He told Threatpost that malware authors are likely to create a library to help bypass Bouncer.
While it would always be safe to assume that Google tests submitted apps on a virtualized phone, Oberheid says by poking around in the virtualized environment, malware authors would be able to work out how to make an app play nice while being tested, and only activate an attack once it’s running on a real phone.
Oberheid’s test showed QEMU as the virtualization environment Google is using.
At the time of their test – since we can probably assume that Google is already revising the test architecture – the researchers have told Forbes the Bouncer test phone is registered to a single user account (Miles.Karlson@gmail.com) and stores two pics (cat.jpg and ladygaga.jpg) as bait for malicious apps to try and steal.