Data Is Another Weapon in the Defence Arsenal

Australia is not only acquiring submarines and missiles as it ramps up its defence capability, but is also investing in data and artificial intelligence.

The country has invested more than AUD500 million over five years in a program called Australia’s One Defence Data — or (OneDD) —  which acts as a unified data lake, providing a secure and central repository that merges fragmented information from disparate sources such as drones, ships and aircraft so the military can make faster and smarter decisions.

Nasa Walton, the chief technology officer at Defence Australia, emphasized the importance of data and AI in national security in a panel discussion at the recent Gartner Summit in Sydney.

“What has been missing in defence is an understanding of how data has become part of military strength and military might,” said Walton.

“It is an understanding of how over the last few years, with the onset of sensors and other information in our capability, data has become a strategic advantage.”

Technology had become so advanced that systems now had only seconds to respond to threats, and Walton contrasted that with the perceived threat of the feared Y2K disaster at the turn of the century.

“We had around 12 months to work towards that, but today we are down to seconds in exploitations of cyber security,” she said.

Walton described OneDD as a “centralized but federated data lake” and said much of the data was “meshed” for specific uses and one-off missions.

“You might call these systems siloed, but think about every time an aircraft goes into flight and creates and needs thousands of data points,” she said.

“I’m not going to re-replicate that into a data lake, but what we do is make sure the right information is given to the right people at the right time, so command and control can make the right decisions.”

Towards private AI

Walton also talked about the evolution of data sovereignty and AI sovereignty, and how – particularly for defence — it needed to move beyond the model of commercial off-the-shelf solutions toward a private approach.

“We need to make sure that we are doing private AI with data we trust and we’re using sovereign,” she said.

“For me, I need to use sovereign because I need trusted, secured information, and I need to be aware of the AI model, and where its training has been and know all the copyright permissions.”

“You need to have your audit logs, your controls, your kill switch, everything designed specifically so that it is going as it was designed,” she added.

“You might call these systems siloed, but think about every time an aircraft goes into flight and creates and needs thousands of data points. I’m not going to re-replicate that into a data lake, but what we do is make sure the right information is given to the right people at the right time, so command and control can make the right decisions.” — Nasa Walton @ Defence Australia

Another panelist, Vinicius Cardoso, said that he had observed the same approach from other organisations, such as those in financial services were security and compliance were also paramount.

"When you bring the models under your control, that can be in your data centre or in your cloud of choice, but it's ring-fenced to your needs in your private network," said Cardoso, who is chief technology officer for Cloudera in Australia and New Zealand.

“A lot of people start with public AI, but it has definitely shifted closer to what Defence has done and adopted a private AI, and no one would disagree that Defence should be one of the most secure organisations in this country.”

Embracing the ‘anti-algorithm’

Walton also spoke on how the rigorous testing of AI models was also reverse engineered using “anti-algorithms.”

“In defence, it’s not just about making sure about the algorithm and the expected outcome,” said Walton. “No, we go back and build the anti-algorithm. Does it give out a different answer than you expect from your positive algorithm?”

“The piece around AI is continual training. You use it, and you bring back results, and you retrain, and you redo, and it is critical for us that we continue to do that,” she added.

Also on the panel was Keir Garrett, the Australian and New Zealand vice president for Cloudera, and she offered some observations on the issue of sovereignty.

“We still have this conversation around the difference between sovereign data and sovereign AI,” said Garrett.

“There is still this kind of this this this thought that because you're using sovereign data that it's automatically sovereign AI and you're protected and it's not.

“We are often visiting customers and having that conversation, and there is a widespread misunderstanding of what that means.”

Image credit: iStockphoto/JasonYht