'AVPass' Sneaks Malware Past Android Antivirus Apps

A team of researchers from Georgia Tech built an Android hacking tool that snuck past nearly all of 58 Android AV antivirus programs in tests conducted via VirusTotal.

Their AVPass toolkit includes a query function that vets and probes an Android AV program's detection capabilities, a malware variant generator that generates multiple variations of a sample, and a data analyzer that analyzes the findings and uses that information to ultimately bypass AV apps on a mobile device.

The researchers plan to release the toolkit at Black Hat USA in Las Vegas next week during their AVPass: Leaking and Bypassing Antivirus Detection Model Automatically session there.

"AVPass is meant to make sure whatever malware you're sending cannot be screened by antivirus," says Max Wolotsky, a PhD student and researcher with Georgia Tech. "The entire goal of AVPass is if you scan malware on either VirusTotal or another AV program" it can't be identified, he says.

Wolotsky and his fellow researchers from Georgia Tech - Chanil Jeon, research associate; Insu Yun, PhD student; Jinho Jung; PhD student; and Taesoo Kim, and assistant professor - say their technique also could work on other platforms, and they plan to test it against Windows desktop machines.

The Python-based AVPass roots out the internal detection methods and code logic of the AV systems information it then uses to cheat the AV system.

Of the dozens of popular and lesser-known AV programs on the free VirusTotal online scanning site, only AhnLab and WhiteArmour's AV programs stopped AVPass in its tracks most of the time, the researchers say.

"We can't say for sure that we can bypass the other 56 AVs 100% of the time; however, in our tests we were almost always able to do so," Wolotsky says. On average, AVPass-generated apps were detected by AV only six percent of the time, he says.

The researchers also learned a few things about Android AV programs in their project: for one thing, the more complex an AV program's detection rules, the stronger its ability to catch malware.

Wolotsky says AV apps can defend against an AVPass-type attack by classifying AVPass as malicious. Android AV app vendors, meanwhile, can rate-limit their AV tools and generate "null" responses so the attack can't glean any intel about the AV program's capabilities.

AVPass sends a series of phony malware variants to test the AV's functions in snippets so as not to release the entire malware sample during the recon phase. With the intel in hand, it then alters the malware. "We found that most AVs commonly use a fixed number of detection rules," Wolotsky says. "For instance, a weak AV can be bypassed only after one feature obfuscation."

The Bigger Picture
The AVPass project is actually just one of multiple research initiatives at Georgia Tech on vulnerabilities in machine learning algorithms. These projects are studying how malicious attackers could manipulate machine learning algorithms and compromise or disrupt security analytics, search engines, customized news feeds, facial and voice recognition, and fraud detection, for example.

Wolotsky and his team's work on AVPass began with exploring how antivirus tools classify malware, and they used VirusTotal to determine what machine-learning techniques the AV programs employ. Their ultimate goal with the project is to find ways for these AV programs to stop malware in its tracks.

AVPass is basically a proof-of-concept tool mainly aimed at developers, both app and AV, that can be used to study ways to detect variations of malware. The Georgia Tech researchers plan to conduct a live demonstration during their Black Hat talk, in which they will submit a piece of their malware to VirusTotal to show how AVPass can be used to determine how to bypass AV systems.

Related Content:

Black Hat USA returns to the fabulous Mandalay Bay in Las Vegas, Nevada, July 22-27, 2017. Click for information on the conference schedule and to register.