Scoring application vulnerabilities
An embodiment includes a method of application vulnerability assessment and prioritization. The method includes ingesting modelling data from data sources for application vulnerabilities. The method includes transforming at least a portion of the modelling data to covariate vectors. The method includes extracting keywords and phrases from the modelling data and statistically measuring relevance of files of the modelling data based on the extracted keywords and phrases. The method includes generating threat levels of the application vulnerabilities based on the covariate vectors and the measured relevance. The method includes outputting the threat levels to a network management system. The method includes implementing, at a first endpoint device of the network, a first patch to address one of the application vulnerabilities.
1. A method of application vulnerability assessment and prioritization, the method comprising:
ingesting modelling data from two or more data sources for a first application vulnerability and a second application vulnerability;
transforming at least a portion of the modelling data to one or more covariate vectors representative of a parameter of the modelling data;
extracting keywords and phrases from the modelling data;
statistically measuring relevance of files of the modelling data based on the extracted keywords and phrases;
generating a first threat level of the first application vulnerability and a second threat level of the second application vulnerability based on the covariate vectors and the measured relevance, wherein the generating includes:
inputting the modelling data, the covariate vectors, and measured relevance into the statistical classification model,
training at least a portion of the statistical classification model on the input to identify one or more features of the modelling data that indicate a threat level posed by the first and the second application vulnerability, and
applying a mathematic model to classified modelling data to calculate a first derived threat level and a second derived threat level;
in response to the first derived threat level differing by more than a threshold amount from the second derived threat level, outputting the first derived threat level as the first threat level and the second derived threat level as the second threat level to a network management system;
causing display of the first threat level and the second threat level in the network management system to enable evaluation of the first application vulnerability and the second first application vulnerability; and
implementing, at a first endpoint device of the network, a first patch to address the first application vulnerability.
2. The method of claim 1 , wherein the modelling data includes one or more or a combination of software weakness data, exploit data, and compromised data.
3. The method of claim 1 , wherein the extracting the keywords and phrases from the modelling data is performed using term frequency-inverse document frequency (TF-IDF) vectorizer.
4. The method of claim 1 , wherein the statistical classification model implements two or more classifiers.
5. The method of claim 4 , wherein a first classifier of the statistical classification model includes an extra trees classifier, and a second classifier of the statistical classification model includes an XGBoost classifier.
6. The method of claim 5 , wherein:
the extra trees classifier is trained based on the input to identify the one or more features; and
the XGBoost classifier is trained on a dataset that only includes the one or more features.
7. The method of claim 1 , wherein the modelling data includes one or more text-based files that include portions of program files and descriptive text related to the first application vulnerability and the second application vulnerability.
8. The method of claim 1 , wherein the covariate vectors are representative of one or more or a combination of an approximately static covariate, an industry-accepted covariate, a simulated compromise covariant, and a derived covariate.
9. The method of claim 1 , further comprising in response to the first derived threat level differing by less than a threshold amount from the second derived threat level:
acquiring environmental data of a computer network implementing a software application affected by the software application vulnerability;
analyzing the environmental data to derive one or more additional features;
reclassifying the modelling data based on the one or more additional features to generate a first reclassified threat level and a second reclassified threat level; and
outputting the first reclassified threat level as the first threat level and the second reclassified threat level as the second threat level to the network management system.
10. The method of claim 1 , wherein the network management system includes a patch management system implemented to distribute patches in the network.
11. The method of claim 1 , wherein:
the covariate vectors are further input to the mathematical model; and
the mathematical model includes one or both of an aggregation of decision trees and conditional logic.
12. The method of claim 1 , wherein:
the two or more data sources include a public site;
the first application vulnerability is associated with a first software application;
the second application vulnerability is associated with a second software application; and
the first endpoint device is implemented in the network and includes the first and the second software applications.
13. A non-transitory computer-readable medium having encoded therein programming code executable by one or more processors to perform or control performance of operations of application vulnerability assessment and prioritization, the operations comprising:
ingesting modelling data from two or more data sources for a first application vulnerability and a second application vulnerability;
transforming at least a portion of the modelling data to one or more covariate vectors representative of a parameter of the modelling data;
extracting keywords and phrases from the modelling data;
statistically measuring relevance of files of the modelling data based on the extracted keywords and phrases;
generating a first threat level of the first application vulnerability and a second threat level of the second application vulnerability based on the covariate vectors and the measured relevance, wherein the generating includes:
inputting the modelling data, the covariate vectors, and measured relevance into the statistical classification model,
training at least a portion of the statistical classification model on the input to identify one or more features of the modelling data that indicate a threat level posed by the first and the second application vulnerability, and
applying a mathematic model to classified modelling data to calculate a first derived threat level and a second derived threat level;
in response to the first derived threat level differing by more than a threshold amount from the second derived threat level, outputting the first derived threat level as the first threat level and the second derived threat level as the second threat level to a network management system;
causing display of the first threat level and the second threat level in the network management system to enable evaluation of the first application vulnerability and the second first application vulnerability; and
implementing, at a first endpoint device of the network, a first patch to address the first application vulnerability.
14. The non-transitory computer-readable medium of claim 13 , wherein the modelling data includes one or more or a combination of software weakness data, exploit data, and compromised data.
15. The non-transitory computer-readable medium of claim 13 , wherein the extracting the keywords and phrases from the modelling data is performed using term frequency-inverse document frequency (TF-IDF) vectorizer.
16. The non-transitory computer-readable medium of claim 13 , wherein:
the statistical classification model implements two or more classifiers;
a first classifier of the statistical classification model includes an extra trees classifier; and
a second classifier of the statistical classification model includes an XGBoost classifier.
17. The non-transitory computer-readable medium of claim 16 , wherein:
the extra trees classifier is trained based on the input to identify the one or more features; and
the XGBoost classifier is trained on a dataset that only includes the one or more features.
18. The non-transitory computer-readable medium of claim 13 , wherein the modelling data includes one or more text-based files that include portions of program files and descriptive text related to the first application vulnerability and the second application vulnerability.
19. The non-transitory computer-readable medium of claim 13 , wherein the covariate vectors are representative of one or more or a combination of an approximately static covariate, an industry-accepted covariate, a simulated compromise covariant, and a derived covariate.
20. The non-transitory computer-readable medium of claim 13 , wherein the operations further comprise:
in response to the first derived threat level differing by less than a threshold amount from the second derived threat level:
acquiring environmental data of a computer network implementing a software application affected by the software application vulnerability;
analyzing the environmental data to derive one or more additional features;
reclassifying the modelling data based on the one or more additional features to generate a first reclassified threat level and a second reclassified threat level; and
outputting the first reclassified threat level as the first threat level and the second reclassified threat level as the second threat level to the network management system.