Microsoft Sentinel UEBA

If you’re an Azure customer and are using Sentinel, then you’ll definitely want to check out Sentinel User and Entity Behavior Analytics (UEBA). It’s an add-on application that works right on top of Sentinel and can be easily setup without any major integrations or customizations. It integrates well with Sentinel SIEM and other Azure security products, allowing you to aggregate your various security use cases and create higher-fidelity alerts.

Sentinel User and Entity Behavior Analytics works by aggregating multiple Azure data sources and finds rarities and outliers within those sources. The data sources are:

  • Azure AD
  • Audit Logs (User and group management activity)
  • Azure Activity (Common Azure operational event logs)
  • Security Events (Windows Security Event Log events)
  • SignIn Logs (Authentication Activity)

The rarities and outliers discovered by Sentinel UEBA are known as “Insights”, which are uncommon actions, devices, peer activity, and other events of interest within your environment. Sample Insights produced are:

  • User performed uncommon action among peers
  • First time user performed this particular action
  • User connected from country uncommonly connected from peer group
  • Unusual number of logon failures performed by user

Sentinel UEBA will automatically produce these Insights based on the activity in your organization, which can flag someone using a new browser, connecting from a new location, or performing an abnormal amount of actions. This type of information can not only help alert on suspicious activity, it can also help an investigator determine events of interest performed by the user when performing an investigation, as many of the Insights would be something an investigator would query manually.

While these can be notable activities, you can also see why these are “Insights” and not “Alerts.” Larger organizations especially will find a lot of this “noisy,” which is why these alone may not be sufficient to justify an investigation without additional context. Behavior Analytics will catch someone still using Internet Explorer, but you may have staff using it to access a legacy application. If you have staff that travel, it will flag their phones connecting to your network as soon as they walk off a plane.

To obtain more value from Behavior Analytics, you can aggregate it with other analytics such as your Sentinel Use Cases, Anomalies, Identity Protection, MCAS, and other Azure alerts. Since the Insight will have the user that performed the action, you can simply aggregate it with your other sources by username.

Most Azure security products will produce an Entity automatically with each alert. For Sentinel use cases, you need to ensure an Entity is set when a rule is configured. For example, your user name field, e.g. AccountName, UserPrincipalName, should be mapped to the Account Entity.

Challenges

Especially for large organizations, Insights will simply be normal activities within your environment. Thus, like with other security products, you may want to filter out those that are common and benign.
Behavior Analytics also provides a priority from zero to ten for each Insight. The least abnormal activities will produce a lower numeric value (0-4), while more rare or uncommon Insights will produce a higher numeric score on the scale (5-10).

If you are going to aggregate Sentinel UEBA with other Azure security products, you may want to explore a weighting or scoring system that ensures Insights don’t outweigh other security alerts. For example, if one user consistently triggers one Insight multiple times, they can repeatedly trigger alerts or appear at the top of your dashboards. Thus, you can cap the score derived from Insights so that it allows other products to equally provide a risk score for the user, so that other high-risk users can generate alerts or appear at the top of dashboards.

Some Insights do not contain numerical values that indicate what baseline and deviations were observed. For example, an Insight that determines an “uncommon number of actions were observed,” the common and uncommon values may not be provided, leaving the analyst unclear if there was a significant deviation from the standard.

Summary

Overall, Sentinel UEBA is a great way to automatically flag suspicious behavior in your environment. Doing the equivalent manually within your organization would require a team of data scientists. Behavior Analytics provides additional criteria you can use to create higher-fidelity alerts, and as well automatically provide investigators will information pertinent to an investigation. If you’re an Azure customer and are using Sentinel then it’s definitely worth checking out.