AI²: an AI-driven predictive cybersecurity platform


In a new paper, researchers from CSAIL and the machine-learning start-up PatternEx have demonstrated an artificial-intelligence platform called “AI²” that can predict 85% of cyber-attacks, by continuously incorporating input from human experts.

To predict attacks, AI² combs through data and detects suspicious activity by clustering the data into meaningful patterns using unsupervised machine-learning. It then presents this activity to human analysts who confirm which events are actual attacks, and incorporates that feedback into its models for the next set of data.


    1. +Diogenes Computer, sensors,routers,bridges, algorithms, learning trees, database, this is all what you need.

  1. So is this a linear regression based algorithm using advanced optimization applied to an incoming threat that has to be acknowledged by the user from the beginning or has it already had training examples pre-loaded prior to running for the first time?
    Or possibly multi class classification neural networking, which looks closer to the multi-layer diagrams in the video.

    1. Melancholia 666 To be fair, i’ve been studying in the field of AI for 3 years now. Pretty sure i’ve fot more experience with machine learning than you do.

    2. @Master Yoda I can suggest you a term “Meta-Learning” if you haven’t known about it.
      It has been 3 years but I am with Dellitsni on this subject. Machine Learning is what enables AI (or “self-consciousness”, the one you expect).
      AI nowadays is a fancy word but yes, as simple as a matrix calculation is still a unit in a process of constructing an AI system. I hope that @Melancholia 666, after few years, has realized it instead of going around and mocking others.

    3. @Long Nguyen-Vu tnx,. for term, did not know about it in name of AI.
      i bet you have more knowledge then me in this field, but my logic, cannot agree that we have “artificial intelligence” which can work without human and theirs input data.

    4. @Master Yoda oh yeah because you have a slightly different definition of “AI”. I totally understand what you’re saying. But this is how researchers in the field coined this term since the dawn of AI. Your expectation is somewhat more about “Artificial General Intelligence” (AGI) which I have referred to the term “Meta-Learning” (you will like it if you read it because it is a way to enable advanced AI). But basically when you talk to people working with machine learning (I am not interested in marketing term), this is the definition of AI that you would hear most of the time. Peace.

    1. This is a very new and fancy department of Artificial Intelligence called Machine Learning. Machine Learning, abbreviated ML, is the act of teaching a computer to perform various tasks, or letting the computer learn by itself (reinforcement learning). Examples of this can be giving your AI script an Atari game – for instance, space shooters – but not telling it how to play. All you tell it is that the further it gets in the level, the better it is. It then “learns”, by training itself for hours upon end, and will, if the machine learning model and script is any good, at some point get better than most human players.

  2. So my question is if AI will eliminate all the threats in cyber security before they will happen why do we still need/ train people in cyber security?

  3. AI is a great supplemental and complementary tool for cybersecurity efforts. Where AI fails to comprehensively analyze data and apply knowledge, humans will fill the gaps.

Leave a Reply

Your email address will not be published. Required fields are marked *

© 2024 EVERYTHING CHATGPT - WordPress Theme by WPEnjoy