An AI created by scientists at the Massachusetts Institute of Technology (MIT) uses machine learning to detect suspicious activity - getting it right 85 per cent of the time.
The system uses an algorithm called "AI2", that detects anomalies, in conjunction with a human expert, because AI2 on its own can lead to false positives, according to MIT News.
"The more attacks the system detects, the more analyst feedback it receives, which, in turn, improves the accuracy of future predictions," said one of the researchers behind the project, Kalyan Veeramachaneni. "That human-machine interaction creates a beautiful, cascading effect."
The merging of artificial intelligence and what researchers call "analyst intuition" has allowed for this new system to be successful in its early development, Veeramachaneni and fellow scientist Ignacio Arnaldo said. They are working on the project alongside a chief data scientist at PatternEx.
The researchers processed three billion pieces of data known as log lines through the system over three months, with the sheer volume of information training the system to the point where it could predict and detect 85 per cent of cyber attacks.
The AI2 algorithm combs through data and presents its findings to a human analyst, who tells it which are real cyber attacks and which are not. It then uses this feedback to refine its analysis of the next set of data.
"You can think about the system as a virtual analyst," Veeramachaneni said. "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."
