University of Michigan researchers have shown that sound waves can be used to hack into devices that use a commonly deployed piece of silicon called a MEMS accelerometer. Fitbits, smartphones, and a variety of medical devices and GPS locators all rely on accelerometers.
The bad news is that the sound-wave hack can be used to control an emerging class of autonomous devices such as drones, self-driving cars, and anything attached to the Internet of Things. The good news: The hack requires physical proximity, expertise in both mechanical and electrical engineering, and above-average programming skills, the researchers tell Dark Reading.
They also admit the actual threat is slight. "We're not saying the sky is falling," says Tim Trippel, one of the researchers and a PhD candidate in the computer science and engineering department at the University of Michigan. "But we need to think about software security and how the hardware can be stimulated environmentally with sound waves and [electromagnetic interference]. If attackers can craft the right type of vibration, they can make a device behave the way they way want it to."
Trippel says the research builds on a paper presented at a 2015 USENIX conference that showed how an acoustical blast could register on a drone as a gust of wind. But rather than just interfere with the accelerometer, the Michigan researchers are taking this to the next level with command-and-control capability.
To demonstrate the acoustical hack, the researchers played a YouTube video from a smartphone that prompted the phone to spell out the word walnut. "We laced a music video with the tones, demonstrating that the interference remains effective even when combined with videos and music that could be automatically played from websites, email attachments, Twitter links tapped on a smartphone," the researchers say in their paper, which will be presented at an IEEE security conference next month in Paris.
The Michigan researchers were also able to use audio tones to disrupt a Fitbit device, artificially adding steps to the device's daily tally. "We also took it a step further to see if we could steer a vehicle, which we did with a toy car," Trippel says.
But infosec professionals don't need to rush out and download a patch or swap out hardware. "This exploit is academic in nature and presents no real-world risk," says Mike Murray, VP of security research and response at mobile security vendor Lookout. "Accelerometer data isn't usually used for any significantly risky purpose."
Trippel says the acoustic attack requires more sophistication than just flooding a network with server requests like a distributed denial-of-service attack, for example. "Attackers would need some knowledge of the algorithms of the sensor data and the signal they're trying to spoof," he adds. "And they'd need physical proximity as well."
The researchers also alerted five chipmakers whose sensors they tested and found vulnerable: Analog Devices, Bosch, InvenSense, Murata Manufacturing, and STMicroelectronics. While acknowledging there's no fundamental flaw, Trippel would like to see the manufacturers alert component customers, who tend to automatically trust sensor data. "It's good hardware that does what it was designed to do," Trippel says, but manufacturers need to make customers who buy them aware so they know when they might fail.
"We're not trying to say all these devices are broken. But going into the age of autonomous systems, we need to be security-aware with hardware and software, and the information fed to those algorithms," Trippel says. With growing reliance on sensors to collect data for industry and consumers, the vulnerability needs to be addressed.
Terry Sweeney is a Los Angeles-based writer and editor who has covered technology, networking, and security for more than 20 years. He was part of the team that started Dark Reading and has been a contributor to The Washington Post, Crain's New York Business, Red Herring, ... View Full Bio