A team of researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has developed a cyber security system entailing artificial intelligence algorithms that can detect up to 85% cyber crimes successfully. The platform has been given the name AI2, Artificial Intelligence squared.
The detection can not only help in predicting attacks beforehand but can also prevent the cyber crime all together. Last week in the New York city at IEEE International Conference on Big Data Security, the team presented a paper under the title “AI2: Training a big data machine to defend”.
As the name implies, the squared means that not only machine is not involved in determining the nature of attack but the human researchers also supplement the decision.
Once the machine suspects the untoward activity, it presents it to the human researchers also known as Analyst Intuition (AI), who confirm the activity as a potential threat. The output is again fed to the algorithm which updates its detection algorithm, thus improving it to a new level of accuracy.
The system reviews 3.6 billion log files daily via machine-learning techniques, declare the suspicious activities as malicious. After processing, the results are presented to human analyst at the end of the day.
The analyst then determines that which attacks are correct and which are false. This feedback is fed back to the system which is used to improve the algorithm for future detections. Thus, there has been an achievement and the false positive rate has improved 5%.
Cyber security is a major concern for the corporations and agencies who are securing their data from cyber criminals who are discovering new ways to intrude at an extremely fast pace.
Here AI2 can prove to be a good guard as Nitesh Chawla, computer science professor at Notre Dame University, describes that AI2 is fast and significant in improving its future prediction process because in few hours it can improve via detecting more attacks and thus receiving more human feedbacks. He says that it is a beautiful interaction between machine and human.