Six Keys to Successful Identity Assurance Strategy: Anomaly Detection
- April 10, 2017
In granting access to users, understanding their behavior goes a long way towards providing frictionless security. As part of our blog series, Six Keys to a Successful Identity Assurance Strategy, we continue to explore going beyond simple two-factor (2FA) or multi-factor authentication (MFA) to create a successful identity assurance strategy for your organization. Previously, we discussed the fundamentals of using business context as part of an effective authentication strategy.
While the attributes and static rules that give us business context provide some ability to assure the right authentication requirements for each situation, watching behavior to see what is normal and what is not can unleash broader capabilities to both improve user experience and security. This is why anomaly detection is the next key to crafting a successful identity assurance strategy.
Identifying Abnormal Access Requests
Let’s start by taking a look at what makes an access request appear abnormal. There’s a simple question we can ask to recognize abnormal: Is this access request unlikely to be legitimate? Answering this question can require information from multiple sources. However, your multi-factor authentication solution should have capabilities to perform a basic level of anomaly detection. Let’s take a look at some examples of these capabilities.
For capabilities such as these that are built into the identity system, we should expect that policies leverage this information to properly deny access or require additional authentication when recognized. We will discuss utilizing other tools to build even further intelligence into recognizing abnormal behavior in a future blog in this series where we focus on getting information from a broader set of systems.
Recognizing Normal through Patterns
In addition to recognizing abnormal behavior, we need to look at the inverse. We must understand normal behavior and how that can impact identity assurance of a user. It all starts with the context attributes we discussed earlier but, in this case, we don’t pre-determine what we trust through static rules. We look at the user and attributes such as their device, location, network, time of day, and access patterns. For each of these, we determine if we’ve seen a common pattern of successful authentication attempts where these attributes are consistent. When we can recognize enough consistency, we can gain some assurance of who this user is without interactive authentication. By using many of the same attributes as we use in business context, looking for patterns instead of static rules, we can start to create intelligence in our identity and access management system.
Putting It All Together
Anomaly detection starts with recognizing situations that could heighten the risk an access request is legitimate. By having the right tools to take appropriate action on this abnormal activity, you build on the business context of an identity assurance strategy.
I include normal behavior in this topic as it makes sense to talk about the abnormal behavior seen with isolating anomalies with recognizing normal patterns. Identifying normal patterns leads into our next blog topic, which is machine learning. The data science behind machine learning becomes important in recognizing normal behavior and can also be leveraged for abnormal behavior. We’ll explore that more in the next blog. In the meantime, download this white paper to learn more about identity assurance.