Scenario: Configuring Adaptive Anomaly Control rules in the Smart mode

17 May 2024

ID 231503

Configuring the Adaptive Anomaly Control rules proceeds in stages:

  1. Training

    After you enable Adaptive Anomaly Control, its rules are in the "Smart training" state. During the training, Adaptive Anomaly Control monitors detections made by the rule and sends detection events to the server.

    If a rule is not triggered at all on a certain device during the training, Adaptive Anomaly Control considers the actions associated with this rule as non-typical. Kaspersky Endpoint Security Cloud will block all actions associated with that rule on that device.

    If a rule is triggered during the training, Kaspersky Endpoint Security Cloud adds events to the detections report and to the Detections of Adaptive Anomaly Control rules repository of the Quarantine.

  2. Processing the list of detections

    Analyze the list of detections in the Detections of Adaptive Anomaly Control rules repository. For each detection, perform one of the following actions:

    • If the detection is not anomalous, add it to exclusions. As a result, this detection and all detections of the same object on other devices are removed from the list. Later, this object will not be detected again on any of your users' devices.

      You can add up to 1000 exclusions for all rules.

    • If the detection is actually anomalous, confirm it. As a result, the detection is removed from the list. Later, if this object is detected again on the same or any other device, it will re-appear in the list of detections.

Each rule has its own training duration that is set by Kaspersky experts. Normally, the training lasts two weeks. The training time is counted separately for each device and only when Kaspersky Endpoint Security for Windows is working on the device. For example, if the training on a device has lasted for a week, and then the device is turned off during a month, the second training week starts only when the device is turned on again.

The training for a rule on a device ends when there are no unprocessed detections over the training duration. That is why we recommend that you process detections at least once a week.

Did you find this article helpful?
What can we do better?
Thank you for your feedback! You're helping us improve.
Thank you for your feedback! You're helping us improve.