PatternEx FAQs

The Intersection of Artificial Intelligence and Information Security

Over time, we have seen a certain number of  questions come up with regularity. If you have a question about how Artificial Intelligence works in InfoSec that is not listed below, please ask by completing the form below.

How does PatternEx use Active Learning?

PatternEx incorporates special learning algorithms that interactively query the user (or some other information source) to obtain feedback at new data points.  

There are situations in which unlabeled data is abundant but labeling that data manually is expensive. In such a scenario, learning algorithms can actively query the user/analyst for labels. This type of iterative supervised learning is called active learning which helps the PatternEx solution to improve the accuracy of its threat detection.

How can PatternEx mimic an InfoSec analyst?

The analyst’s judgment and intuition is represented by the labels the analyst gives the PatternEx Virtual Analyst Platform. 

The AI2 technology has a facility that takes those labels, then analyzes exactly the relationships across all the dimensions in the underlying data that caused the human to label a behavior malicious or not. From this analysis, PatternEx synthesizes a predictive model. This model can make predictions of analyst labels on incoming events in real time. These predictions are validated by the analyst, and the feedback is absorbed by the platform again.

Over time, this prediction-correction feedback loop becomes extremely accurate. Eventually, the PatternEx Platform “mimics” the analyst by predicting exactly what the analyst would label a given behavior.

What is a rare event modeler?

An ensemble of unsupervised machine learning algorithms that are combined in a way to create a model that best represents the underlying data. The output of the rare event modeler is prediction that an observed event is rare. Such events are passed to the analyst for labels.

What is the difference between supervised and unsupervised learning?

Supervised learning is the machine learning task of inferring a function from labeled training data. Unsupervised learning is the machine learning task of inferring a function to describe hidden structure from unlabeled data. Since the examples given to the learner are unlabeled, there is no error or reward signal to evaluate a potential solution. This distinguishes unsupervised learning from supervised learning. Read more on the PatternEx blog.

Ask the Experts