Sunday, February 9, 2025

How A Generative AI UEBA Solution Can Power Your Defenses

 


I have written a lot about Generative AI in the past, both in books, at my full-time job, my freelancing gig, and on this blog site.  As mentioned, it brings both its good and bad sides with it.  But to stay positive today (despite what else is going on in the news), a great big advantage that Generative AI is its ability to harness through tons of data and provide the appropriate response to a query.  So, in this regard, it can be a great boon for Cybersecurity as well.

One  instance of this is filtering through all the noise that are outputted in terms of the log files from the network security devices that you may have implemented.  For example, this can include firewalls, network intrusion devices, routers, hubs, etc. 

They all present information that is especially useful to an IT Security team.  But the problem with this is that there are tons of it comb through.  It can take an IT Security team days and even months to have go through all of this. 

But by using Generative AI and professionally training it, the model can sift through all of this very quickly, in fact in just a matter of a few minutes.  From here, it can then present the information that is relevant to the IT Security team.  Once notable example of this is for the filtering of what are known as “False Positives”.  These are the alerts and warnings that come through that are deemed to be illegitimate, or exceptionally negligible risk.

A good model can detect all of this, and either completely discard them or archive them for later study.  From here, only the real alerts and  warnings are then presented to the IT Security team, which can then be triaged and responded to appropriately. 

This almost eliminates the problem of what is known as “Alert Fatigue”.  This is where there are so many of them to go through, one can get burned out.  But this can cause severe repercussions as well, as a burned-out employee could decide to totally give up, and not even respond to anything.

Another key area where Generative AI can be used in quite well is in a field called “User and Entity Behavioral Analytics”, also known as “UEBA” for short.  It can be technically defined as follows:

“User and entity behavior analytics (UEBA) is a cybersecurity solution that uses algorithms and machine learning to detect anomalies in the behavior of not only the users in a corporate network but also the routers, servers, and endpoints in that network.”

(SOURCE:  What is User Entity and Behavior Analytics (UEBA)? | Fortinet)

Deploying and using this kind of solution can be quite complex, depending upon how large your IT and Network Infrastructure is, and how many employees that you have.  But simply put, UEBA is the science of tracking down any abnormal or unusual patterns in the usual flow of network traffic. 

A scenario where is used most is in trying to determine any anomalies that fall outside of the baseline profile that you have established.  A great use case is when you have set up a limit of only three login attempts after an account is locked out.

An outlier here is if somebody keeps trying repeatedly to login.  Usually, this is a warning sign  that a Cyberattacker is on the prowl by launching a Dictionary Attack, or it can be a frustrated employee that is legitimately trying to login into their device.  Whatever the situation might be, this must be investigated.  But once again, having to go through all the data can be a nightmare.  But by incorporating Generative into your solution, any suspicious behavior that merits further attention by the IT Security team will be presented very quickly by the model.

In fact, many of the Cybersecurity vendors are already baking this into their solutions, so there is no extra work that is required on your end.  All you need to do is merely feed the data so the model can learn and create a baseline profile.  Once it has done this, you then need to set up the criteria of what is deemed to be abnormal behavior.

UEBA solutions are used quite heavily in Security Operation Centers (also known as “SOCs”), because of all the monitoring that must be 24 X 7 X 365.  But some of the areas where it has hardly ever been deployed are the following:

*The Healthcare Industry

*Government Agencies

*The Educational Sector

The first and last ones are hit the hardest by the Cyberattacker, because legacy systems are still being used, and lack of funding, especially by the schools.  But the good thing here is that most of the UEBA solutions are now offered as a SaaS based product, which makes it affordable for about any kind of entity. 

It is highly likely that UEBA will develop over time, especially as Generative AI quickly advances further.  Thus, if you decide to deploy it, you will have to make sure that you deploy all the software patches and upgrades to it.

My Thoughts on This:

Making use of a UEBA solution is of course a no brainer.  It is one of the best ways that you can defend your business from an Insider Threat.  It also comes very handy when trying to secure those login credentials that are deemed to be “super user” (this falls under the realm of Privileged Access Management).  Consider these stats:

*In 2024, the cost of a Data Exfiltration Attack rose from $4.4 billion to well over $4.8 billion, which is a 10% increase.

*Over 70% of SOCs feel that they will miss a real threat with all the False Positives that are constantly being bombarded with.

(SOURCE:  Behavioral Analytics in Cybersecurity: Who Benefits Most?)

But also, there are two key things to keep in mind:

*The baseline profile that you get is only going to be as accurate as the data you feed into the model for it to learn.

*While using an automated tool by Generative AI is advantageous, do not become overly dependent upon it.  Remember, great Cybersecurity takes an equal combination of both technology and the human element.

But best of all, a good UEBA solution will reduce “Alert Fatigue”, and help ensure a sense of proactiveness amongst your IT Security team.

No comments:

Post a Comment

A Fine Line Must Be Drawn In Generative AI Usage: The Banking Example

  One common question that I get asked from time to time is what do Cyberattackers like to prey on?   In other words, who do they like to ta...