MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL), along with researchers at security firm PatternEx, have revealed a new system THERE (Vision market or Artificial Intelligence) called AI2, and can detect the 85% of cyber attacks, with false positives five times smaller than all existing solutions.
The new MIT system is not entirely based on artificial intelligence (AI), but also on user control, which researchers call an analyst intuition or AI, hence the name of the AI system2 or AI in the square.
Η study (PDF) AI2: Training a big data machine to defend was presented at the IEEE Big Data Security International Conference last week in New York.
Researchers report that they have fed AI2 with over 3,6 billion lines of logs, allowing the system to crawl content using techniques engineeringof learning.
At the end of each day, the system presents its findings to an administrator (sysadmin), which then confirms or rejects the security alerts.
This human feedback is then incorporated into the AI learning system2 and is used the next day to analyze new data.
After tests carried out by MIT and PatternEx researchers, they reported that AI2 catches 85% accuracy in detecting cyber attacks, which is 2,92 times better than the success rates of similar cyber-attack automated detection systems in use today.
In addition, the rate of false positives was also lower, five times smaller than that achieved by similar cyber-security solutions.
The impressive for AI2 is that, over time, the AI system becomes smarter and can recognize more and more attack agents, and the person who controls the system will not be necessary, as at the beginning.