Demand for graphics processing units, or GPUs, has exploded in recent years as video rendering and artificial intelligence systems have increased the need for processing power. And while most of the most visible shortages (and rising stock prices) affect the top-end PC and server chips, mobile graphics processors are the version that everyone with a smartphone uses every day. So vulnerabilities in these chips or in their implementation can have real-world consequences. This is precisely why Google’s Red Team to find vulnerabilities in Android has targeted open-source software from chip giant Qualcomm, which is often used to implement mobile GPUs.
At the Defcon security conference in Las Vegas on Friday, three Google researchers presented more than nine — now patched — vulnerabilities they discovered in Qualcomm’s Adreno GPU, a software suite used to coordinate between GPUs and an operating system such as Android on Qualcomm phones. Such “drivers” are critical to the design of a computer and have broad permissions in an operating system’s kernel to ensure coordination between hardware peripherals and software. Attackers could exploit the vulnerabilities discovered by the researchers to gain complete control of a device.
For years, engineers and attackers alike have focused on potential vulnerabilities in a computer’s central processing unit (CPU) and optimizing the efficiency of GPUs, relying on their raw processing power. But as GPUs become increasingly important to everything a device does all the time, hackers on both ends of the spectrum are exploring how GPU infrastructure could be exploited.
“Compared to the large Android ecosystem, we are a small team. The scope is too big for us to cover everything. So we need to find out what has the biggest impact,” says Xuan Xing, manager of Google’s Android Red Team. “Why did we focus on a GPU driver in this case? Because untrusted apps don’t need permission to access GPU drivers. This is very important and I think will attract the attention of many attackers.”
Xing is referring to the fact that applications on Android phones can communicate directly with the Adreno GPU driver, “without sandboxing, without additional permission checks,” as he puts it. This doesn’t give applications the ability to act uncontrolled in and of itself, but it makes GPU drivers a bridge between the regular parts of the operating system (where data and access are carefully controlled) and the system kernel, which has full control over the entire device, including its memory. “GPU drivers have all kinds of powerful features,” Xing says. “This mapping in memory is a powerful primitive that attackers want.”
The researchers say the vulnerabilities they uncovered all stem from the intricacies and complicated connections GPU drivers must handle to coordinate everything. To exploit the vulnerabilities, attackers would first have to gain access to a target device, perhaps by tricking victims into downloading malicious apps.
“There are a lot of moving parts and no access restrictions, so GPU drivers are easily accessible to virtually any application,” says Eugene Rodionov, technical lead of the Android Red Team. “What makes things really problematic here is the complexity of the implementation – this is one point that is responsible for a number of vulnerabilities.”
Qualcomm has issued patches for the vulnerabilities to original equipment manufacturers (OEMs) that use Qualcomm chips and software in the Android phones they make. “Regarding the GPU issues disclosed by the Android Security Red Team, patches were made available to OEMs in May 2024,” a Qualcomm spokesperson tells WIRED. “We recommend that end users install security updates from device manufacturers as they become available.”
The Android ecosystem is complex, and patches must be pushed from a vendor like Qualcomm to OEMs, and then packaged by each individual device maker and delivered to users’ phones. This trickle-down process sometimes means devices are left unprotected, but Google has invested years to improve these pipelines and streamline communication.
Nevertheless, the results provide further evidence that GPUs themselves and the software that supports them have the potential to become a crucial battleground in computer security.
Rodionov puts it this way: “The combination of high complexity of implementation and wide accessibility makes it a very interesting target for attackers.”
