In response to the latest MITRE Engenuity ATT&CK® Evaluation 3, McAfee noted five capabilities that are must-haves for Sec Ops and displayed in the evaluation. This blog will speak to the alert actionability capability which is essential. This critical ability to react in the fastest possible way, as early as possible on the attack chain, while correlating, aggregating and summarizing all subsequent activity while reducing alert fatigue to allow Sec Ops to uphold efficient actionability.
As a Sec Ops practitioner and former analyst, I can remember the days of painstakingly sifting through countless alerts to determine if any of them could be classified as an incident. It was up to me to decide if the alert were a false positive, false alarm, or something the business should take more seriously… was it something we should wake someone up in the middle of the night over?
It’s been years since I sat on the front line, triaging the results of millions of dollars in investments installed on 100’s of 1000’s of systems worldwide. Thank goodness, times have changed. But the concept of “Alert Actionability” is still a very real aspect of SOC tooling, and it seeks to address 3 primary factors: trustworthiness, detail, and reaction capabilities.
When I say “trustworthiness” I’m referring to a quality of fidelity that has two equal, yet opposing, faces of efficacy: false positives and false negatives. Now, it would be very easy for a SOC solution provider to claim that its product offers 100% visibility if it creates an alert for every process activity and artifact recorded. Sure, its coverage is present, but how actionable is the needle in a stack of needle? As a result, the vendor is likely pressured to fine tune it’s alerting and as such introduces the risk of false negatives, or actual malicious events which go undetected. In the zeal of appealing to useability requirements the false positive curve decreases but the false negative volumes have no choice but to rise.
Resulting in a graph like this:
The secret sauce in the vendor’s capabilities lies in its capacity to push the intersection of these as far right as possible: minimize the false positives and maximize true positives while simultaneously attempting to bring false negatives down to zero. The better a vendor’s product can perform these non-trivial goals, the more likely it is to win your trust as a solution! And the more likely you are to trust the results you see on the dashboard.
Endpoint Detection and Response (EDR) tools have a unique property in which they offer both telemetry and alerting. This implies that there are two goals for EDR platforms: to include event level (telemetry) visibility with automated detection and to provide alerting capabilities for triggering action and triage. With telemetry, the concept of “falsing” is negated because it’s used in a post-facto context. After the alert is constructed, the telemetry can be correlated with the alert logic to provide supporting details. Simply, for EDR telemetry, the more the better.
As an analyst, I remember how much I loved putting together the pieces to tell a story. Extracting key artifacts from several disparate data sources and correlating hypothesis allowed me to present a compelling case as to the conclusion of the alert’s disposition. And I knew that I needed as much detail as possible to make my case; this is just as true today. The detail needs to be easily accessible, and it’s even better when the platform provides the detail proactively. In cases where such supporting evidence may not be possible in the alerting, an analyst’s expectation is that the platform makes hunting for those details easy; I’d even venture to say, “a delight.”
Many EDR platforms on the market offer reaction capabilities to address the “Response” moniker of the acronym. How flexible those response capabilities are in the platform provides a domain of options to act in response to the alert. For example, its rather evident that once an alert is convicted, the analyst may want to block the process, or remove a file from disk. But these reactions imply that the conviction is monolithic in that the analyst is absolutely sure of her conclusion. What if the conclusion is that we simply need more data? Having a robust reaction library that allows for further investigation with routines like sending a sample to a running sandbox, interacting with a given endpoint to act as an administrator, view system logs, or check the history of network connections all empower the analyst with further investigatory options. But why stop there? Having any fixed set of reactions would be presumptive. Instead, EDR products with a dynamic library and flexible, customizable, and modular reaction platform is key as every single SOC I’ve ever worked with has unique Incident Management and Standard Operating Procedures.
MITRE ENGINUITY™ released results for its 3rd round of ATT&CK® Evaluations in April 2021. The industry is certainly fortunate to receive such 3rd party efficacy testing in the EDR market completely free to consumers. It is incredibly important to add that the ATT&CK Evaluations should be used as a single component of your EDR evaluation program. Efficacy helps determine how fit-for-purpose the product is by answering questions like, “Will it detect a threat when I need it to?” or “Can I find what I need, when I need it?”. But practitioners realize there are also pivotal points that need to be addressed around manageability. Understanding that not alerting on everything is just as important as alerting on the right things. And giving you a plethora of alerting response capabilities helps complete the alert investigation and response actions. McAfee’s MVISION EDR embraces all of these key alert actionability factors and will help displace the manual efforts in your analytics processes. McAfee’s MVISION EDR (soon to evolve to MVISION Extended Detection & Response (XDR)) provided insight through detail and reduced alert fatigue during the evaluation providing context and enrichment, resulting in a ratio of 62% analytic detections (non-telemetry detections) out of the 274-total detections.
Check out other McAfee discussion on MITRE (see resources tab.)