New proof-of-concept tool detects stealthy malware hiding in graphics cards

As anti-virus solutions become more robust and Microsoft becomes better at plugging Windows vulnerabilities, malware designers have to get more creative about attacking PCs and servers. One wide-open avenue of attack: hardware components like graphics and network cards. Yes, you read that right.

Security software isn’t necessarily looking for malware lurking within peripherals, opening the door for Bad Guys to stash nasty code in your killer, pixel-pumping Radeon or GeForce graphics card. But fear not! Berlin-based researcher Patrick Stewin says he has figured out a way to detect this tricky malware without taxing the CPU, according to SC Magazine Australia.

Peripheral-based malware is particularly difficult to detect, because it doesn’t have to rely on a weakness in your computer’s operating system, according to Stewin. Instead, it takes advantage of the processing power already present in peripherals like graphics cards that may not be expecting an attack.

Graphics cards, sound cards, and other PC components can process data using direct memory access (DMA). Instead waiting to receive data processing via a PC’s CPU, a graphics card can bypass the CPU to access and process graphical data directly from memory.

DMA helps make a PC work faster and reduces the load on the CPU. But it also means that a properly designed bit of malware can get in through a data-crunching peripheral. Once infected, DMA attacks can do all sorts of damage, such as copying encryption keys or installing other types of malware for identity theft, though the odds of being infected by this advanced type of malware are admittedly slim.

Comparing notes

Stewin’s solution–dubbed BARM–deals with DMA attacks by keeping tabs on what a system is supposed to be doing and comparing that to what the PC actually is doing.

Stewin took a piece of DMA malware called DAGGER (which SC Magazine says was created by Stewin and another security researcher) and inserted it into a PC. Then Stewin used BARM to monitor the data flowing through the PC’s memory system.

By analyzing what the computer should be doing based on what the user was asking of it, then comparing that activity to the data that was actually running through the PC’s memory, Stewin could scan for anomalies that could be malware.

At this point, BARM is only a proof-of-concept piece of software. And it’s easy to imagine software like this setting off false positives if BARM’s expectations for what your PC should be doing are incorrect. Nevertheless, the hope is that BARM could one day constantly monitor a PC for DMA-based attacks without taking a big bite out of the computer’s processing power.

Stewin’s work will be presented during the International Symposium on Research in Attacks, Intrusions and Defenses in October.