cannot be detected and quarantined by traditional means. What is the best way, then, to find the GPU malware and mitigate this risk?

Anywhere there is firmware that can be updated, there is the possibility for implanting malware. Most antimalware tools do not check the firmware for all of the devices on a system to determine if it has been compromised. It may not even be possible to analyze the firmware on an endpoint to determine if it has been compromised.

Graham Cluley blogs about new proof-of-concept malware, from Team Jellyfish, that hides in the GPU. Currently this GPU malware only works on Linux, but it could be applied to Windows and OSX in the future. Analyzing the endpoint directly won’t necessarily detect the GPU malware, but if it uses the network connection of the compromised host, this network traffic could be detected. If network traffic is detected without an identified running process, that could be an indicator of malware buried deep inside the endpoint. The initial code executed to load the malware in the GPU could be detected, as could any files stored on the file system that were used in the attack.

For enterprises with high security requirements, mitigating the risk of GPU malware might require the removal of any internal device with firmware, but most likely fully replacing the hardware will be most effective since multiple internal components could have been infected. The firmware on the device could be updated with known, good firmware to clean the malware.

Ask the Expert:
Have a question about enterprise threats? Send it via email today. (All questions are anonymous.)