Many of us are connected to the world through our devices. With the touch of a button, we can adjust the temperature of our homes, turn off the lights, share our daily steps with a friend, and send a string of emojis to our family group chat. Unfortunately, with the convenience of interconnected devices comes the threat of breached communications. This is a reality not only for everyday consumers, but for our government, industries, military, and other sectors.

With an increase in vulnerability and threats to secure communications, experts in the field of communications have been working to improve security and limit the number of threats to our system. In an effort to tackle the growing demand for secure communications research and development, the Intelligence Advanced Research Projects Activity has awarded a $14 million contract to fund a collaborative project between BAE Systems and a team of researchers at Virginia Tech.

The goal of the award is to develop tools to decipher an ever-growing number of radio frequency signals in an effort to quickly and accurately help secure mission-critical information. Of that $14 million, the team at Virginia Tech is receiving a nearly $1.5 million sub-award to provide expertise in the area of machine learning based strategies for radio frequency anomaly detection.

Lingjia Liu, a professor in the Bradley Department of Electrical and Computer Engineering, serves as the principal investigator of the sub-award. Faculty members Jeff Reed, Carl Dietrich, and Harpreet Dhillion, also of the department, are serving as co-principal investigators of the project.

Liu is working on machine learning-based spectrum prediction. Specifically, he and his team will focus on “reservoir computing.” This type of computing is used to predict activity and occupancy of an entire network based on observations of a small sample piece of that network. Looking at a small section to predict activity on a larger scale is particularly important when it comes to secure communications because analyzing the entire network in a time-sensitive situation would be nearly impossible. Because information travels quickly, threats can cause damage on a wide scale in just seconds. Being able to identify a threat as quickly as possible is key to preventing damage on a large scale.

This computing method is also known as the recurrent neural network. The team will also use signal characterization to identify the types of signals being sent within the secure communications network. With these prediction and characterization techniques, the hope is that the technology produced will provide enhanced situational awareness, help target threats, and secure communications against malicious attacks.

Shashank Jere, a Ph.D. candidate studying machine learning and wireless networks at Virginia Tech, also is working on the project and is excited learn and contribute to research with such a high impact. Jere will be contributing to the development of the artificial intelligence-based methods used to detect the occurrence of those “anomalous” or abnormal signals.

“Anomalous signals could be any wireless activity other than standard signals such as Wi-Fi, Bluetooth, or LTE/5G cellular signals,” said Jere. “Such a framework would be the foundation toward preventing jamming or spoofing attacks in existing potentially sensitive wireless networks.”

Jere said the applications of these research findings don’t necessarily stop with military communications. Security and privacy are becoming increasingly important to consumers, especially as more of daily life become intertwined with hand-held devices. Having quick access to bank accounts, medical information, and GPS location are all features that consumers find convenient, but that sensitive information could be targeted in data breaches.

“The outcomes from this research could be ported to the consumer wireless device industry, where security and privacy are becoming increasingly important considerations,” Jere said.

By the end of the three-year project, Liu expects that real applications from this research should be in place to help mitigate the number of threats and malicious attacks to secure communications.

When it comes to partnering with industry collaborators such as BAE Systems, Liu said,  “Partnering with leading companies on this kind of project helps us make our research real and relevant.”

The “real and relevant” work, he said, comes from using actual data sets obtained from testbeds instead of simulated data obtained through equations and simulations. When working with industry, he said, the research and tasks usually have clear objectives and timelines for each stage of the project. This approach is very different from working with other universities, which usually have open-ended research.

“Getting instantaneous feedback from industry leaders in the field is a great way to gain relevant and practical knowledge,” said Liu.

BAE will provide the team with suggestions and guidance on how to set up the experiment. This assistance includes providing a baseline of simulated data before moving on to actual data obtained from the hardware testbed.

The advanced defense technology company will then review the success of the proposed candidate technologies (and how well they worked to analyze anomalies, threats, etc.) together with the Virginia Tech team and provide suggestions and feedback based on those test runs.

As part of the ongoing research project, Liu and his team have bi-weekly meetings with the prime contractor, BAE Systems. The Virginia Tech team of researchers is responsible for developing the machine learning algorithms based on the simulated data. BAE will then incorporate that design and the algorithms into its software tools.

Nima Mohammadi, another Virginia Tech graduate research assistant on the team, described his excitement to be working on a project of this scale.

“This is an excellent opportunity to work on a multidisciplinary project that allows us to leap from theory to practice and make an attempt at overcoming an exciting and tangible – but of course challenging – problem which could really push the edge of science,” said Mohammadi. “Interdisciplinary studies are exactly where most of the innovations take place, and this venture could yield many options for a Ph.D. student in both academia and industry.”