SCAN ENGINE AUTOSCALING USING CLUSTER-BASED PREDICTION MODELS
Various embodiments include systems and methods to implement predictive scan autoscaling using cluster-based prediction models by a security platform to predict scanning loads associated with computing resources. Predictive scan autoscaling using cluster-based prediction models may improve the security posture of computing resources by improving the speed by which a security platform may scan for threats of a cyberattack. The security platform may predict scanning loads based on data indicative of previous scanning loads over one or more periods of time for clusters of similar client networks, where similarity may be based on a comparison of deployment assets. The security platform may combine predicted scanning loads with requests for scans received from various client networks.
1 .- 20 . (canceled)
21 . A method comprising:
performing, by one or more computer systems that implement a security service:
performing scans of a plurality of client networks to assess the client networks for security vulnerabilities or security events;
building, based on observation data about a cluster of the client networks, a prediction model that predicts a security service load of security service associated with the client networks;
determining, based on the prediction model, a first number of compute instances to use to perform scans on the cluster at a particular time in future;
deploying, prior to the particular time, the first number of compute instances at the security service; and
using, at the particular time, the first number of compute instances to perform scans on the cluster.
22 . The method of claim 21 , wherein the prediction model is a machine learning model trained using one or more machine learning techniques.
23 . The method of claim 21 , wherein the prediction model is a trend model that was fitted to a time series data.
24 . The method of claim 23 , wherein the trend mode comprises an auto-regressive integrated moving average (ARIMA) model.
25 . The method of claim 21 , wherein the cluster of the client networks is determined using a clustering algorithm, based on similarities in asset deployments of the client networks.
26 . The method of claim 25 , wherein the similarities are determined using snapshot data collected from the client networks.
27 . The method of claim 26 , wherein the snapshot data is collected by agents of the security service executing within the client networks.
28 . The method of claim 26 , wherein the prediction model is updated periodically based on newly collected snapshot data.
29 . The method of claim 21 , wherein the security service is implemented in a cloud computing environment provided by a cloud service provider, and the compute instances are part of a pool of virtual machine instances.
30 . The method of claim 21 , further comprising the security service:
scaling up and down a pool of compute instances based on repeated predictions of the prediction model.
31 . A system comprising:
a security service implemented using one or more computer systems, configured to:
perform scans of a plurality of client networks to assess the client networks for security vulnerabilities or security events;
build, based on observation data about a cluster of the client networks, a prediction model that predicts a security service load of security service associated with the client networks;
determine, based on the prediction model, a first number of compute instances to use to perform scans on the cluster at a particular time in future;
deploy, prior to the particular time, the first number of compute instances at the security service; and
use, at the particular time, the first number of compute instances to perform scans on the cluster.
32 . The system of claim 31 , wherein the prediction model is a trend model that was fitted to a time series data.
33 . The system of claim 32 , wherein the trend mode comprises an auto-regressive integrated moving average (ARIMA) model.
34 . The system of claim 31 , wherein the security service is implemented in a cloud computing environment provided by a cloud service provider, and the compute instances are part of a pool of container instances.
35 . The system of claim 31 , wherein the cluster of client networks are located at a particular geographic region, and the compute instances are deployed at the particular geographic region.
36 . The system of claim 31 , wherein the cluster of the client networks is determined using a clustering algorithm, based on similarities in asset deployments of the client networks.
37 . The system of claim 31 , wherein the observation data comprises observed scan request to the security service and observed performance metrics of the security service.
38 . The system of claim 37 , wherein:
the security service queues scan requests from the client networks when there are insufficient number of computing instances to handle the scan requests; and
the performance metrics include a number of queued scan requests.
39 . The system of claim 31 , wherein the security service is configured to:
scale up and down a pool of compute instances based on repeated predictions of the prediction model.
40 . The system of claim 39 , wherein:
the security service is configured to scale up and down the pool according to a regular period; and
the regular period is based on (a) a particular day of a week, month, or year, or (b) a particular time of a day.