You are here
$1.5M Wireless/Machine Learning DARPA Award for Device Fingerprinting
ECE Professor Kaushik Chowdhury is leading an interdisciplinary COE team leveraging a $1.5 million DARPA grant to identify device-specific radio signals on a massive scale.
Every individual device in the global wireless Internet of Things—estimated to reach over 20 billion devices by 2020—communicates via radio transmitters and receivers. Even when multiple devices are transmitting the same information, each device imprints its own unique signal pattern on the transmission that makes it possible to identify individual smartphones or laptops.
“Due to manufacturing variances, each electronic device has minor hardware differences in its processing chain, which make it slightly different from every other device,” explains Kaushik Chowdhury, an associate professor in Electrical and Computer Engineering at Northeastern. “You can think of these distinct characteristics as a fingerprint. By studying the unique properties of a received radio signal, you can identify which device is sending it.”
Identifying the source of a signal has important implications for both cyber security and emergency preparedness. By impersonating the signal of an authorized device, cyber terrorists could disrupt the safe operation of an airplane or access critical national security data. In the realm of emergency preparedness, rapidly identifying nearby devices could aid in alerting emergency responders, saving crucial seconds and potentially saving human lives.
Based on these and other important implications of device fingerprinting, the US Defense Advanced Research Projects Agency (DARPA) recently awarded $1.5 million in funding to an interdisciplinary team in Northeastern’s College of Engineering (COE) working to optimize this capability. The team will apply machine-learning algorithms to make signal identification practical on a massive scale.
Chowdhury is joined on this team by fellow COE faculty members Stratis Ioannidis, Tommaso Melodia, and Jennifer Dy. All are professors in the ECE Department, but they bring a range of interdisciplinary expertise to this complex challenge. While Chowdhury and Melodia focus their research on wireless networks and communications, Ioannidis and Dy have expertise in signal processing. Their shared goal is to adapt machine-learning techniques such as deep convolutional neural networks—which are already proven for image recognition—into the domain of wireless signal pattern recognition.
The collaborative team will leverage the DARPA funding to develop new methodologies and machine-learning architectures that can correctly classify 10,000 devices with an accuracy rate of 99 percent. The researchers will use a 14-TeraByte database of radio transmitters and radio frequency (RF) signals collected by DARPA across different wireless channels to demonstrate the robust operation of their innovative new methods.
According to Chowdhury, the Northeastern team represents the only all-academic effort to address this extremely complex technical problem posed by DARPA. “That’s not a coincidence,” Chowdhury notes. “The College of Engineering at Northeastern has worked hard to develop leading expertise in the Internet of Things and related topic areas, recognizing the huge importance of the IoT to society, industry, and the military. It’s gratifying to see all this expertise coming together to solve a practical problem and achieve a common goal.”
Chowdhury also notes the important role played by Northeastern’s Integrated Science and Engineering Complex (ISEC) in supporting this kind of collaborative, interdisciplinary research. “The faculty and graduate students involved in this project all reside together on two floors on the ISEC building,” he says. “That physical proximity facilities formal joint research, as well as interesting hallway conversations that can fuel real innovation.”
“By making aggressive investments in supporting interdisciplinary research, Northeastern’s College of Engineering is now uniquely poised to solve today’s most pressing and complex engineering challenges, like device fingerprinting—which has never before been attempted at such a massive scale. I’m excited to be part of what’s happening here,” concludes Chowdhury.