UC San Diego Scientists Lay the Groundwork to Catch “Rogue” Wireless Privacy and Security Threats

Wireless signals help us achieve results large and small—from scheduling appointments to defending national security. To many of us, these invisible, airborne communications form a constantly shifting mesh that structures our daily lives. But not all of these unseen signals are helpful. 

Experts in the fields of wireless communications and signal processing are on the front lines of trying to detect and isolate “rogue” wireless signals that pose real risks to privacy and security. To aid these efforts in the area of national defense, researchers with the UC San Diego Qualcomm Institute (QI) and Jacobs School of Engineering’s Department of Electrical and Computer Engineering have introduced two novel techniques that use a combination of machine learning and classical methods to accomplish an important step in separating rogue signals from their benign counterparts. 

“This ongoing work has important implications for wireless security,” said Srivatsan Rajagopal, a postdoctoral researcher in the lab of QI affiliate and Department of Electrical and Computer Engineering faculty member Dinesh Bharadia. “I am quite encouraged by our papers’ initial promising results and look forward to seeing their potential impact on the industry.”

Rajagopal is first author on the paper “Blind Signal Characterization: Transformers, Triplet Losses and beyond” and a co-author on the study “Fourier Meets Gardner: Robust Blind Waveform Characterization,” led by Electrical and Computer Engineering graduate student Radhika Mathuria.

Rajagopal and Mathuria presented the papers at the IEEE International Symposium on Dynamic Spectrum Access Networks in Washington, D.C., in early May.

How to Spot a Trojan Horse

Rogue signals are not always easy to tell apart from their legitimate counterparts. Sometimes, these “Trojan horse” signals closely resemble those that follow industry standards. This allows rogue signals to sneak undetected through security measures in standard wireless receivers and help bad actors access critical information on the other side. 

Often, these signals arrive without any identifying information, including their point of origin, their frequency or other parameters that programmers might use to identify them. 

Rajagopal and Mathuria set out to create novel algorithms that would set the groundwork for this research. Their algorithms automatically process incoming signals’ mode of delivery, from a single frequency band to multiple frequency bands, the first step toward separating friend from foe.

The researchers began by using Searchlight, a pre-processing system introduced in an earlier study, to cluster incoming signals for the new algorithm to analyze. In both papers, the team created its own, robust dataset, which it used as the basis for simulated tests in one paper, and both simulated and real-world tests in the second.

Rajagopal and Mathuria also tackled estimating an incoming signal’s properties, like its frequency, through signal processing techniques, toward an end goal of better describing a signal through its characteristics. 

The algorithms introduced in both papers fill a gap in current decoders, which assume that the receiver already knows a signal’s frequency and other characteristics. 

This work is supported in part by the Office of the Director of National Intelligence’s Intelligence Advanced Research Projects Activity.