A Canberra-based scientist is taking the guesswork out of border security by teaching computers to spot suspicious activity.
Dr Aditya Menon is one of the lead scientists on the initiative which pairs CSIRO's Data61 research team with global IT company Unisys.
"The aim of the international collaboration with Unisys is to develop machine-learning algorithms to detect suspicious records in cargo and passenger data," Dr Menon said.
"These algorithms work by sifting through large volumes of historical data to automatically identify what constitutes a suspicious record.
"Once trained in this manner, these algorithms offer an automated way to determine the risk profile of cargo or passengers."
If successful, the algorithms would help border security staff quickly identify high-risk passengers and freight that needed to be checked, he said.
"This lets security experts focus their efforts only on the most relevant high-risk cases, rather than spend time unnecessarily on cases that can be easily cleared.
"The work is important because the sheer volume of cargo shipments and passenger travel is such that security experts cannot possibly go through each case individually."
The research could be employed to detect security threats posed by travellers, visa applicants and cargo, and will soon be trialled at a major freight hub in Asia.
"Border security is a top issue for many governments globally, and Unisys and Data61 have the capability to help many governments across the world," Dr Menon said.
Machine learning is an analytical technique which uses algorithms to teach computers to get better in the way they use data.
"For example, we may have historical data on cargo shipments that were flagged as suspicious, but we'd expect there to be a few suspicious shipments that weren't flagged," Dr Menon explained.
"These pose a problem when trying to learn from historical data as to what characterises a suspicious shipment, and my research looks at how we might solve this."
The director of border security at Unisys, John Kendall, said the technology would help frontline staff become more efficient.
"This will allow border agencies to automate the processing of low risk people and cargo while reserving specialised border security resources for the small percentage of travellers and caro that present a higher risk profile."