System and method for machine learning model determination and malware identification
A system and method for batched, supervised, in-situ machine learning classifier retraining for malware identification and model heterogeneity. The method produces a parent classifier model in one location and providing it to one or more in-situ retraining system or systems in a different location or locations, adjudicates the class determination of the parent classifier over the plurality of the samples evaluated by the in-situ retraining system or systems, determines a minimum number of adjudicated samples required to initiate the in-situ retraining process, creates a new training and test set using samples from one or more in-situ systems, blends a feature vector representation of the in-situ training and test sets with a feature vector representation of the parent training and test sets, conducts machine learning over the blended training set, evaluates the new and parent models using the blended test set and additional unlabeled samples, and elects whether to replace the parent classifier with the retrained version.
1. A method comprising:
receiving information indicating a first plurality of attributes associated with a first plurality of files for machine learning training;
determining, based on the information, a plurality of classifications for a second plurality of files;
based on adjudicating one or more of the plurality of classifications, determining a second plurality of attributes associated with the second plurality of files;
determining, based on training and using at least a portion of the first plurality of attributes and at least a portion of the second plurality of attributes, a machine learning model; and
determining, based on the machine learning model, that a file comprises malicious content.
2. The method of claim 1 , wherein the first plurality of attributes was extracted from the first plurality of files, wherein the first plurality of attributes comprises at least one of a file header property, a component of a file, or a binary sequence, and
wherein the second plurality of attributes was extracted from the second plurality of files wherein the second plurality of attributes comprises at least one of a file header property, a component of a file, or a binary sequence.
3. The method of claim 1 , wherein the training is based on a machine learning algorithm, and wherein the machine learning algorithm comprises a decision tree, support vector machine, a k-nearest neighbor algorithm, an artificial neural network, or a Bayesian network.
4. The method of claim 1 , wherein each classification of the plurality of classifications comprises at least one of malicious content or benign content.
5. The method of claim 1 , wherein the adjudicating comprises:
confirming classifications of the plurality of classifications that are correct; and
adjusting classifications of the plurality of classifications that are not correct.
6. The method of claim 1 , wherein the adjudicating indicates an accuracy of the machine learning model at determining the plurality of classifications and relevant attributes for determining the plurality of classifications.
7. The method of claim 1 , wherein the training is triggered based on adjudicating a threshold number of classifications of the plurality of classifications.
8. The method of claim 1 , wherein the second plurality of files are unique to an organization.
9. A device comprising:
one or more processors; and
memory storing instructions that, when executed by the one or more processors, cause the device to:
receive information indicating a first plurality of attributes associated with a first plurality of files for machine learning training;
determine, based on the information, a plurality of classifications for a second plurality of files;
based on adjudicating one or more of the plurality of classifications, determine a second plurality of attributes associated with the second plurality of files;
determine, based on training and using at least a portion of the first plurality of attributes and at least a portion of the second plurality of attributes, a machine learning model; and
determine, based on the machine learning model, that a file comprises malicious content.
10. The device of claim 9 , wherein the first plurality of attributes was extracted from the first plurality of files, wherein the first plurality of attributes comprises at least one of a file header property, a component of a file, or a binary sequence, and
wherein the second plurality of attributes was extracted from the second plurality of files wherein the second plurality of attributes comprises at least one of a file header property, a component of a file, or a binary sequence.
11. The device of claim 9 , wherein the training is based on a machine learning algorithm, and wherein the machine learning algorithm comprises a decision tree, support vector machine, a k-nearest neighbor algorithm, an artificial neural network, or a Bayesian network.
12. The device of claim 9 , wherein each classification of the plurality of classifications comprises at least one of malicious content or benign content.
13. The device of claim 9 , wherein the adjudicating comprises:
confirming classifications of the plurality of classifications that are correct; and
adjusting classifications of the plurality of classifications that are not correct.
14. The device of claim 9 , wherein the adjudicating indicates an accuracy of the machine learning model at determining the plurality of classifications and relevant attributes for determining the plurality of classifications.
15. The device of claim 9 , wherein the training is triggered based on adjudicating a threshold number of classifications of the plurality of classifications.
16. The device of claim 9 , wherein the second plurality of files are unique to an organization.
17. A non-transitory computer-readable medium storing instructions that, when executed, cause:
receiving information indicating a first plurality of attributes associated with a first plurality of files for machine learning training;
determining, based on the information, a plurality of classifications for a second plurality of files;
based on adjudicating one or more of the plurality of classifications, determining a second plurality of attributes associated with the second plurality of files;
determining, based on training and using at least a portion of the first plurality of attributes and at least a portion of the second plurality of attributes, a machine learning model; and
determining, based on the machine learning model, that a file comprises malicious content.
18. The non-transitory computer-readable medium of claim 17 , wherein the first plurality of attributes was extracted from the first plurality of files, wherein the first plurality of attributes comprises at least one of a file header property, a component of a file, or a binary sequence, and
wherein the second plurality of attributes was extracted from the second plurality of files wherein the second plurality of attributes comprises at least one of a file header property, a component of a file or a binary sequence.
19. The non-transitory computer-readable medium of claim 17 , wherein the adjudicating comprises:
confirming classifications of the plurality of classifications that are correct; and
adjusting classifications of the plurality of classifications that are not correct.
20. The non-transitory computer-readable medium of claim 17 , wherein the adjudicating indicates an accuracy of the machine learning model at determining the plurality of classifications and relevant attributes for determining the plurality of classifications.
21. A system comprising:
a first computing device configured to:
receive information indicating a first plurality of attributes associated with a first plurality of files for machine learning training;
determine, based on the information, a plurality of classifications for a second plurality of files;
based on adjudicating one or more of the plurality of classifications, determine a second plurality of attributes associated with the second plurality of files;
determine, based on training and using at least a portion of the first plurality of attributes and at least a portion of the second plurality of attributes, a machine learning model; and
determine, based on the machine learning model, that a file comprises malicious content; and
a second computing device configured to:
send, to the first computing device, the information.
22. The system of claim 21 , wherein the first plurality of attributes was extracted from the first plurality of files, wherein the first plurality of attributes comprises at least one of a file header property, a component of a file, or a binary sequence, and
wherein the second plurality of attributes was extracted from the second plurality of files wherein the second plurality of attributes comprises at least one of a file header property, a component of a file, or a binary sequence.
23. The system of claim 21 , wherein the training is based on a machine learning algorithm, and wherein the machine learning algorithm comprises a decision tree, support vector machine, a k-nearest neighbor algorithm, an artificial neural network, or a Bayesian network.
24. The system of claim 21 , wherein each classification of the plurality of classifications comprises at least one of malicious content or benign content.
25. The system of claim 21 , wherein the adjudicating comprises:
confirming classifications of the plurality of classifications that are correct; and
adjusting classifications of the plurality of classifications that are not correct.
26. The system of claim 21 , wherein the adjudicating indicates an accuracy of the machine learning model at determining the plurality of classifications and relevant attributes for determining the plurality of classifications.
27. The system of claim 21 , wherein the training is triggered based on adjudicating a threshold number of classifications of the plurality of classifications.
28. The system of claim 21 , wherein the second plurality of files are unique to an organization.