Researchers at MIT are developing a platform for electronic predicted using artificial intelligence attacks

Apr

19

2016

Developed by researchers from the Massachusetts Institute of Technology MIT in collaboration with emerging Technologies machine learning a new platform to anticipate cyber attacks by using artificial intelligence.

It is noteworthy that the current security systems are divided into two categories: human or mechanism. And rely on human solutions to the rules set by the security experts so he is unable to fend off attacks that do not coincide with those rules.

While based approach relies on machine learning to “detect anomalies”, which tends to raise the so-called “false positives”, which creates a lack of trust in the system and ends up being a refuge for human achievement.

The new platform, bearing the name of the AI 2, which developed in the “laboratory of computer science and artificial intelligence” of the Institute of MIT and PatternEx company specialized in machine learning, to combine the two approaches together.

The group and presented in his paper, the platform characterized by its ability to anticipate cyber attacks much better than existing systems through a Continuously integrating inputs from experts mortal.

The team demonstrated the ability of AI platform 2 to detect 85% of the attacks, which is better than the previous standards by about three times, and was also able to reduce the “false positive results” One factor of 5.

The platform has been tested on a 3.6 billion piece of known as the log data lines Log Lines, which has been created by millions of users over the three months.

And to predict the attacks, looking AI 2 in the data reveal any suspicious activity by assembling data into meaningful patterns used techniques to machine learning is monitored. Then submit this activity to analysts mortal mission Altakd of events that are the actual attacks, to incorporate these results later in the models for adoption at the next set of data.

Experts say it is difficult to create a security-mail systems integrate human and automated approaches, partly because of the challenge description data security-mail to the algorithms manually.

For example, experts say that for the development of an algorithm for computing vision able to identify objects with high accuracy, the description of their own data is simple. It’s enough to run a few human volunteers to label images either “bodies” or “other”, and then enter this data into the algorithm.

But for the security of mail, we can not resort to the ordinary who do not have the skill necessary to label and categorize attacks people. It takes relevant experts.

This is based, according to experts, another problem, namely: security experts are busy, then he can not afford to spend all day in a review of packets of data that have been reported as suspicious.

The team says that the AI 2 is able to schedule the billions of log lines per day, and convert the data cut – minute basis by – minute to the “features” are different, or separate types of behavior which is considered the ultimate “natural” or “unnatural.”

Viewing:-130

In: Technology & Gadgets Asked By: [15174 Red Star Level]

Answer this Question

You must be Logged In to post an Answer.

Not a member yet? Sign Up Now »