Smart software and strategic partnerships to keep our radio waves safe

Finding ‘anomalies’ in the increasing cacophony of signals and voices traversing our radio waves is the task a trio of researchers at the University of Melbourne have set themselves. Their aim: to help protect the radio spectrum and the technologies that rely on it, from malicious interference.

Professor Tansu Alpcan from the Department of Electrical and Electronic Engineering is leading the project with Professor Chris Leckie and Dr Sarah Monazam Erfani both from the School of Computing and Information Systems.

In May this year, they were awarded an Australian Research Council (ARC) Linkage Grant for the project. Leading global security and defence company, Northrop Grumman Corporation (NGC), is their industry partner; Justin Kopacz is NGC’s partner investigator.

Professor Alpcan says the radio spectrum is a national asset and one that is increasingly relied on to control critical infrastructure, including the use of wireless networks and Internet-of-Things (IoT) applications for communication, commerce, and security.

One incident that highlights the vulnerability of radio frequencies is the 2016 case of a Melbourne man who pretended to be an air traffic controller and made hoax calls to pilots flying into Melbourne Airport. He used cheap, off-the-shelf radio equipment and made calls over two days, including a fake pilot-in-distress mayday broadcast.

These fake calls could have caused a major accident endangering hundreds of innocent lives, says Professor Alpcan. While this caller was found and stopped quickly, another operator who used different strategies and who was more intent on causing harm could have been harder to find.

As the technologies and systems using radio frequencies become more complicated, such as 5G, he says new threats and vulnerabilities are emerging in both commercial and defence operations.

Our project looks at how we can more quickly identify these anomalies or unauthorised broadcasts with possible ill intent, using a distributed radio network.

By analysing the data from across the radio spectrum, they can provide ‘situational awareness’, helping operators to identify and make sense of the unexpected signals so that appropriate action can be taken.

The ARC project will use modern machine and deep learning algorithms to do this, processing data from networks of software defined radio (SDR) nodes.

SDR equipment is now widely available and inexpensive. It is capable of operating across a wide range of radio frequencies, rather than the fixed frequencies coded into many common such radios, televisions, mobile phones and those enabled with Wi-Fi or Bluetooth capabilities.

It is the very flexibility of SDR that creates new vulnerabilities.

Professor Alpcan says while a detected anomaly could be a ‘friendly’ transmission, a distress call from a lost hiker, for instance, it could also be from an adversary.

The machine learning algorithms will help to sort the signals that need further attention from the overall noise of the radio-frequency spectrum. They will also be trained to identify attacks intent on deceiving machine learning algorithms themselves.

Professor Alpcan says this is an exciting part of the project, which will draw on game theory to develop adversarial machine learning approaches, to train and strengthen algorithms against efforts to deceive them.

The ARC project is focused on basic research. However, Professor Alpcan expects that, with Northrop Grumman as an industry partner, it will lead to practical commercial and defence cybersecurity applications to protect the radio-based elements of new technologies.