Could we finally be on the verge of realizing the promise of artificial intelligence (AI)? Speaking at Slush 2019, Mikko Hypponen, F-Secure’s Chief Research Officer, revealed that significant leaps in how AI is used are possible now by applying some inhuman thinking to machine learning.
Mikko recalled reading an article in a Finnish technology magazine that suggested the time for AI-like technology was “just around the corner.”
Those words were published in April of 1983.
Today, machine learning and artificial intelligence are progressing faster than ever. Yet most of the discussion about AI still tends to revolve around what Mikko called “the stereotype” of “superhuman intelligence”—computers that think or act like humans, but in superhuman ways.
“But maybe that’s not the right way to think about it,” he said. “Maybe we can do better.”
Project Blackfin is a new project from F-Secure. It aims to rethink the possibilities of artificial intelligence by applying collective intelligence that could mimic superhuman abilities demonstrated by the cooperative behavior of animals, like schools of fish, flocks of birds, or ant colonies.
Mikko suggested that we could tap into this “swarm intelligence” to improve how we handle on of the bigger problems on earth—cyber security.
A pretty big problem
“When we do our job right, it’s invisible,” Mikko said.
He compared doing computer security at a global scale to the classic video game Tetris.
“Your successes disappear, but your failures pile up.”
Unfortunately, the opportunities for failure are almost infinite. Cyber security pros “have to fight massive amounts of attacks, millions of malware samples, millions of attacks.”
Fighting cyber crime on this scale in a way that is seamless for people and businesses creates a massive technological challenge.
Mikko quoted Gustav Söderström, Spotify’s Chief R&D Officer, who said at Slush, “You have to have a pretty big problem for artificial intelligence to make sense as solution.”
Computer security, Mikko noted, is “a pretty big problem.”
The wisdom of the swarm
F-Secure first applied machine learning to cyber security in 2005.
“We didn’t call it ‘artificial intelligence’ because we are Finnish,” Mikko said. “We called it something much more boring.”
The way this company and cyber security companies in general have typically used machine learning is to correlate and cross-reference data collected from vast numbers of endpoints.
“When we use these new swarm intelligence features, and when we add intelligence within one agent, more intricate behaviors and more complex capabilities emerge.”
He points out that the uses of this technology could be useful in multiple fields. These include logistics, energy, and—of course—cyber security.
“And then we add in to agent-based learning things like federated-learning mechanisms. Then it’s not just about the intelligence of one individual agent.”
With network connectivity, the agents could exchange data to improve decision-making. If the agents can connect out to the cloud, this the data shared can further enhance the swarm’s effectiveness. But internet access – which can be unavailable for several reasons, including a malware attack — isn’t necessary.
“With federated learning mechanisms, it works better with network connectivity. But it doesn’t need it, “ Mikko said. “Alone is good. Together is even better.”
The future is now
F-Secure is already using swarm intelligence in our Rapid Detection & Response Service.
“We already stopped mechanisms from reporting false alarms,” Mikko explained.
Given the massive shortages in cyber security professionals and the ceaseless parade of increasingly advanced attacks, anything that helps focus resources to only real and pressing issues gives businesses an advantage.
“It already works. And this is the very beginning of a multi-year project.”
Want to know even more about Project Blackfin?
Check out how this post on how we got here by Matti Aksela, who leads our Artificial Intelligence Center of Excellence.