It’s very useful to know when there will be a cyber attack before it happens. Now MIT has a program that can predict attacks in 85% of cases.
Such programs work in two ways: some are seeking artificial intelligence anomalies in Internet traffic. It works, but it happens to display and wrong results and find a threat where none exists. Other software systems are built on the basis of guidelines developed by people, but it’s hard to create a system to capture any risk of attack.
MIT artificial intelligence has built a new, suggestively named AI2, which combines the two approaches. AI2 uses three different learning algorithms to detect abnormal events.
Like any artificial intelligence system, however, it needs feedback from a man to determine whether those events are suspicious or even if it’s an error. And obviously, as ordinary users could not differentiate between various attacks, the first sets of results would be forwarded to experts, according to the Gizmodo.
In tests conducted to date, MIT artificial intelligence data used 3.6 billion related to Internet activity. It could accurately identify 85% of attacks, timely, and gave five times fewer false alarms than other similar artificial intelligence. The results were presented at the International Conference on Data Security in New York.
Experts argue that the program will become even more efficient. “The more attacks detected, the expert receives more feedback and, in turn, will improve the accuracy of future forecasts, explained Kalyan Veeramachaneni.