Systems and methods for training malware classifiers
The disclosed computer-implemented method for training malware classifiers may include (1) perturbing, at a computing device, a binary file in a manner that maintains functionality of the binary file, (2) classifying the perturbed binary file with a first machine learning classifier to produce a classification result, (3) producing a transformed file by repeating the perturbing and classifying steps until the transformed file becomes misclassified, and (4) performing a security action comprising training a second machine learning classifier with the transformed file and an associated correct classification result. Various other methods, systems, and computer-readable media are also disclosed.
1. A computer-implemented method for training malware classifiers, at least a portion of the method being performed by a computing device comprising at least one processor, the method comprising:
classifying, at the computing device, a binary file as benign;
perturbing, at the computing device, the binary file in a manner that maintains functionality of the binary file;
classifying, at the computing device, the perturbed binary file with a first machine learning classifier to produce a classification result;
producing, at the computing device, a transformed file by repeating the perturbing and classifying steps until the transformed file becomes classified as malware, the repeated perturbing and classifying comprising:
identifying at least one functionally equivalent instruction for at least one instruction included in an instruction-level representation of the binary file;
iteratively changing instructions in the binary file by substituting the at least one functionally equivalent instruction for the at least one instruction in the instruction-level representation of the binary file; and
retaining at least a subset of the instruction changes that produce the classification result; and
performing a security action comprising training a second machine learning classifier with the transformed file and an associated classification result.
2. The computer-implemented method of claim 1 , further comprising:
classifying a second binary file with the trained second machine learning classifier to produce a respective classification result for the second binary file; and
performing a second security action in response to the respective classification result for the second binary file.
3. The computer-implemented method of claim 2 , wherein the second security action further comprises blocking access by the second binary file to the computing device, another computing device, or both.
4. The computer-implemented method of claim 1 , wherein the repeated perturbing and classifying further comprises reordering instructions included in the instruction-level representation of the binary file.
5. The computer-implemented method of claim 1 , wherein the repeated perturbing and classifying further comprises removing an instruction from the instruction-level representation of the binary file.
6. A system for training malware classifiers, the system comprising:
at least one physical processor; and
physical memory comprising computer-executable instructions that, when executed by the physical processor, cause the physical processor to:
classify, at the system, a binary file as benign;
perturb, at the system, the binary file in a manner that maintains functionality of the binary file;
classify the perturbed binary file with a first machine learning classifier to produce a classification result;
produce a transformed file by repeating the perturbing and classifying steps until the transformed file becomes classified as malware, the repeated perturbing and classifying comprising:
identifying at least one functionally equivalent instruction for at least one instruction included in an instruction-level representation of the binary file;
iteratively changing instructions in the binary file by substituting the at least one functionally equivalent instruction for the at least one instruction in the instruction-level representation of the binary file; and
retaining at least a subset of the instruction changes that produce the classification result; and
perform a security action comprising training a second machine learning classifier with the transformed file and an associated classification result.
7. The system of claim 6 , wherein the repeated perturbing and classifying further comprises reordering instructions included in the instruction-level representation of the binary file.
8. The system of claim 6 , wherein the repeated perturbing and classifying further comprises removing an instruction from the instruction-level representation of the binary file.
9. The system of claim 6 , wherein the computer-executable instructions further include computer-executable instructions that, when executed by the physical processor, cause the physical processor to:
classify a second binary file with the trained second machine learning classifier to produce a respective classification result for the second binary file; and
perform a second security action in response to the respective classification result for the second binary file.
10. A non-transitory computer-readable medium comprising one or more computer-executable instructions that, when executed by at least one processor of a computing device, cause the computing device to:
classify, at the computing device, a binary file as benign;
perturb, at the computing device, the binary file in a manner that maintains functionality of the binary file;
classify the perturbed binary file with a first machine learning classifier to produce a classification result;
produce a transformed file by repeating the perturbing and classifying steps until the transformed file becomes classified as malware, the repeated perturbing and classifying comprising:
identifying at least one functionally equivalent instruction for at least one instruction included in an instruction-level representation of the binary file;
iteratively changing instructions in the binary file by substituting the at least one functionally equivalent instruction for the at least one instruction in the instruction-level representation of the binary file; and
retaining at least a subset of the instruction changes that produce the classification result; and
perform a security action comprising training a second machine learning classifier with the transformed file and an associated classification result.
11. The non-transitory computer-readable medium of claim 10 , wherein the repeated perturbing and classifying further comprises reordering instructions included in the instruction-level representation of the binary file.
12. The non-transitory computer-readable medium of claim 10 , wherein the repeated perturbing and classifying further comprises removing an instruction from the instruction-level representation of the binary file.
13. The non-transitory computer-readable medium of claim 10 , wherein
training a second machine learning classifier with the transformed file and an associated classification result comprises training the second machine learning classifier with the transformed file and a classification of the transformed file indicating the transformed file is benign.
14. The system of claim 6 , wherein
training a second machine learning classifier with the transformed file and an associated classification result comprises training the second machine learning classifier with the transformed file and a classification of the transformed file indicating the transformed file is benign.
15. The computer-implemented method of claim 1 ,
training a second machine learning classifier with the transformed file and an associated classification result comprises training the second machine learning classifier with the transformed file and a classification of the transformed file indicating the transformed file is benign.
16. The non-transitory computer-readable medium of claim 10 , wherein the one or more computer-executable instructions, when executed by the at least one processor, further cause the computing device to:
classify a second binary file with the trained second machine learning classifier to produce a respective classification result for the second binary file; and
perform a second security action in response to the respective classification result for the second binary file.
17. The computer-implemented method of claim 1 , wherein the repeated perturbing and classifying further comprises re-ordering register-preservation pushes.
18. The computer-implemented method of claim 1 , wherein the repeated perturbing and classifying further comprises swapping registers.
19. The system of claim 6 , wherein the repeated perturbing and classifying further comprises at least one of re-ordering register-preservation pushes or swapping registers.
20. The non-transitory computer-readable medium of claim 10 , wherein the repeated perturbing and classifying further comprises at least one of re-ordering register-preservation pushes or swapping registers.