Researchers from MIT’s Computer Science and
Artificial Intelligence Laboratory (CSAIL) have developed an artificial
intelligence (AI) platform that can ‘predict 85% of cyberattacks’, so long as
it benefits from human input.
Its latest paper, AI2: Training a
big data machine to defend, revealed that this unique approach is
capable of delivering better results than machines or humans would be able to
alone.
The collaborative effort sees the AI system
take the lead – it ‘combs’ through data, highlighting the parts it considers to
be suspicious and characteristic of cyberattacks.
Humans then take over, analysing the information
provided by the system to then validate the findings. Feedback is then
passed back to the system, helping improve AI2’s detection capabilities.
This process continues and with each iteration, its
ability to accurately identify cyberattacks improves.
“You can think about the system as a virtual
analyst,” explained Kalyan Veeramachaneni, a research scientist at CSAIL who
co-developed the system.
“It continuously generates new models that it can
refine in as little as a few hours, meaning it can improve its detection rates
significantly and rapidly.”
What is interesting about these findings is that it
underscores the importance of human input, as a recent Wired
article noted.
Speaking to the publication, Mr. Veeramachaneni explained
that with cybersecurity evolving all the time, non-machine insight is something
that cannot be replicated.
He said: “The attacks are constantly evolving. We
need analysts to keep flagging new types of events. This system doesn’t get rid
of analysts. It just augments them.”