Automating cluster interpretation in security environments
Disclosed herein are methods, systems, and processes to automate cluster interpretation in computing environments to develop targeted remediation security actions. To interpret clusters that are generated by a clustering methodology without subjecting clustered data to classifier-based processing, separation quantifiers that indicate a spread in feature values across clusters are determined and used to discover relative feature importances of features that drive the formation of clusters, permitting a security server to identify features that discriminate between clusters.
1. A computer-implemented method for automating cluster interpretation in a cybersecurity environment, the computer-implemented method comprising:
accessing a dataset generated by a plurality of computing devices;
executing an unsupervised clustering methodology on the dataset;
executing, without using a classifier, a supervised classification methodology, wherein the executing the supervised classification methodology comprises:
determining a separation quantifier that indicates variance between a plurality of features shared by each cluster of a plurality of clusters generated by the unsupervised clustering methodology, wherein each feature of the plurality of features is a security vulnerability of a plurality of security vulnerabilities associated with one or more computing devices of the plurality of computing devices; and
determining a feature importance of each feature of the plurality of features for each cluster of the plurality of clusters; and
configuring a targeted remediation security action for the one or more computing devices based on the feature importance of each feature indicating variance between each cluster of the plurality of clusters.
2. The computer-implemented method of claim 1 , wherein
the dataset comprises a plurality of data points, and
the execution of the unsupervised clustering methodology generates a cluster label for each data point of the plurality of data points.
3. The computer-implemented method of claim 2 , further comprising:
determining a centroid value for each cluster.
4. The computer-implemented method of claim 3 , wherein determining the separation quantifier further comprises:
(a) selecting a first cluster and a first feature of a plurality of features of the first cluster,
(b) defining a base value of the first cluster based on a first centroid value of the first cluster in the first feature,
(c) determining a separation value between the base value and a plurality of centroid values for the first feature in each cluster other than the first cluster, and
(d) determining a relative feature importance of each feature for the first cluster by repeating (a), (b), and (c) for each feature of the plurality of features other the first feature.
5. The computer-implemented method of claim 4 , further comprising:
normalizing the relative feature importance of each feature to determine the feature importance of each feature.
6. The computer-implemented method of claim 5 , wherein determining the separation value, further comprises replacing a mean value of the first cluster with the base value of the first cluster.
7. A non-transitory computer readable storage medium with program instructions executable to automate cluster interpretation in a cybersecurity environment and configure a targeted remediation security action, comprising:
accessing a dataset generated by a plurality of computing devices, wherein
the dataset comprises a plurality of data points, and
a first data point of the plurality of data points is associated with an identity of a computing device of a plurality of identities of the plurality of computing devices;
processing the dataset with an unsupervised clustering methodology;
performing, without using a classifier, a supervised classification methodology, wherein the performing the supervised classification methodology comprises:
determining a separation quantifier that indicates a degree of variance for a plurality of features shared by each cluster of a plurality of clusters generated by the unsupervised clustering methodology, wherein:
determining the separation quantifier comprises determining a standard deviation of each cluster other than a selected cluster of the plurality of clusters by replacing a mean value of the selected cluster with a base value of the selected cluster, and
each feature of the plurality of features is a security vulnerability of a plurality of security vulnerabilities associated with the computing device; and
determining a feature importance of each feature of the plurality of features for each cluster of the plurality of clusters; and
determining whether a selected feature of the plurality of features separates the selected cluster from each cluster other than the selected cluster based on the degree of variance and a corresponding feature importance of the selected feature; and
configuring the targeted remediation security action for the computing device based on the corresponding feature importance enumerated by the degree of variance between the plurality of features shared by the selected cluster and one or more other clusters of the plurality of clusters.
8. The non-transitory computer readable storage medium of claim 7 , wherein at least one feature of the plurality of features is a binary feature.
9. The non-transitory computer readable storage medium of claim 7 , wherein determining the separation quantifier, further comprises:
selecting the selected cluster and the selected feature,
defining the base value of the selected cluster based on a centroid value of the selected cluster in the selected feature,
determining a separation value between the base value and a plurality of centroid values for the selected feature in each cluster other than the selected cluster, and
determining a relative feature importance of the selected feature based on a coefficient measure assigned to the selected feature.
10. The non-transitory computer readable storage medium of claim 7 , wherein configuring the targeted remediation security action comprises:
an unsupervised clustering methodology, and
a supervised classification methodology performed without using a classifier.
11. A system comprising:
one or more processors; and
a memory coupled to the one or more processors, wherein the memory stores program instructions executable by the one or more processors to automate cluster interpretation in a cybersecurity environment and configure a targeted remediation security action by:
accessing a dataset generated by a plurality of computing devices, wherein:
the dataset comprises a plurality of data points, and
a first data point of the plurality of data points is associated with an identity of a computing device of a plurality of identities of the plurality of computing devices;
processing the dataset with an unsupervised clustering methodology;
performing, without using a classifier, a supervised classification methodology, wherein the performing the supervised classification methodology comprises:
determining a separation quantifier that indicates a degree of variance for a plurality of features shared by each cluster of a plurality of clusters generated by the clustering methodology, wherein:
determining the separation quantifier comprises determining a standard deviation of each cluster other than a selected cluster of the plurality of clusters by replacing a mean value of the selected cluster with a base value of the selected cluster, and
each feature of the plurality of features is a security vulnerability of a plurality of security vulnerabilities associated with the computing device; and
determining a feature importance of each feature of the plurality of features for each cluster of the plurality of clusters; and
determining whether a selected feature of the plurality of features separates the selected cluster from each cluster other than the selected cluster based on the degree of variance and a corresponding feature importance of the selected feature; and
configuring the targeted remediation security action for the computing device based on the corresponding feature importance enumerated by the degree of variance between the plurality of features shared by the selected cluster and one or more other clusters of the plurality of clusters.
12. The system of claim 11 , wherein at least one feature of the plurality of features is a binary feature.
13. The system of claim 11 , wherein determining the separation quantifier, further comprises:
selecting the selected cluster and a selected feature of the plurality of features of the selected cluster,
defining the base value of the selected cluster based on a centroid value of the selected cluster in the selected feature,
determining a separation value between the base value and a plurality of centroid values for the selected feature in each cluster other than the selected cluster, and
determining a relative feature importance of the selected feature based on a coefficient measure assigned to the selected feature.
14. The system of claim 11 , further comprising:
determining the centroid value for the selected cluster, wherein the base value of the selected cluster is the centroid value of the selected cluster in the selected feature.
15. The computer-implemented method of claim 1 , wherein the unsupervised clustering methodology comprises at least one of:
K-Means clustering;
Agglomerative clustering;
Mean-Shift clustering;
Density-based spatial clustering; or
Expectation Maximization clustering.
16. The computer-implemented method of claim 1 , wherein the dataset comprises at least one of:
dangerous user behavior data;
threat probing data;
threat movement data;
remote entry data;
failed access attempt data;
malicious behavior on asset level data;
suspicious behavior on asset level data;
malicious behavior on network level data; or
account adjustment data.
17. The non-transitory computer readable storage medium of claim 7 , wherein the unsupervised clustering methodology comprises at least one of:
K-Means clustering;
Agglomerative clustering;
Mean-Shift clustering;
Density-based spatial clustering; or
Expectation Maximization clustering.
18. The non-transitory computer readable storage medium of claim 7 , wherein the dataset comprises at least one of:
dangerous user behavior data;
threat probing data;
threat movement data;
remote entry data;
failed access attempt data;
malicious behavior on asset level data;
suspicious behavior on asset level data;
malicious behavior on network level data; or
account adjustment data.
19. The system of claim 11 , wherein the unsupervised clustering methodology comprises at least one of:
K-Means clustering;
Agglomerative clustering;
Mean-Shift clustering;
Density-based spatial clustering; or
Expectation Maximization clustering.
20. The system of claim 11 , wherein the dataset comprises at least one of:
dangerous user behavior data;
threat probing data;
threat movement data;
remote entry data;
failed access attempt data;
malicious behavior on asset level data;
suspicious behavior on asset level data;
malicious behavior on network level data; or
account adjustment data.