Machine learning through iterative memory analysis for malware detection
A system and method of anti-malware analysis including iterative techniques. These techniques are used to create a file attribute tree used by a machine learning analyzer to identify malicious files.
1. A computer-implemented method for training a machine-learning static analyzer based on iterative executable file analysis, the method executed on a processor, the method comprising the steps of:
a. retrieving file attributes from an untrusted file;
b. creating an attributes tree of the untrusted file's attributes;
c. performing iteration of dynamic analysis of the untrusted file including decryption of at least one encrypted data block of the untrusted file by intercepting an operation of decrypting the at least one encrypted data block during an execution of the untrusted file's code and updating the tree of the file attributes for the untrusted file based on the at least one decrypted data block;
d. repeating step (c) until the file's code has been executed;
e. training machine learning static analyzer based on the updated file attributes tree of the untrusted file and a classified files collection;
f. analyzing the untrusted the and a second untrusted file, wherein the second untrusted file was modified from the untrusted file after the execution; and
g. repeating the steps of intercepting and updating at least three times, wherein updating at least three times expands the file attributes tree with six additional attributes for analysis.
2. The method of claim 1 , further comprising the step of decrypting data blocks in new or existing memory.
3. The method of claim 2 , further comprising the step of freezing the execution of the file following analysis of additional file data.
4. The method of claim 3 , further comprising the step of decrypting data blocks of the untrusted file in new or existing memory.
5. The method of claim 1 , further comprising making a given area executable and switching execution to the given area.
6. A non-transitory computer-readable storage medium with program code for training a machine-learning static analyzer based on iterative executable file analysis by way of the following steps comprising:
a. retrieving file attributes from an untrusted file;
b. creating an attributes tree of the file attributes of the untrusted file;
c. performing iteration of dynamic analysis of the untrusted the by:
intercepting an operation of decrypting at least one encrypted data block during an execution of the untrusted file's code; and
updating the tree of the file attributes for the untrusted file based on the at least one decrypted data block;
d. repeating step (c) until the file's code has been executed;
e. training the machine learning static analyzer based on the updated file attributes tree of the untrusted file and a classified files collection;
f. analyzing the untrusted file and a second untrusted file, wherein the second untrusted file was modified from the untrusted file after the execution; and
g. repeating the steps of intercepting and updating at least three times, wherein updating at least three times expands the file attributes tree with six additional attributes for analysis.
7. The non-transitory computer readable storage medium of claim 6 , further comprising decrypting data blocks of the untrusted file in new or existing memory.
8. The non-transitory computer-readable storage medium of claim 7 , further comprising the step of freezing the execution of the untrusted the following analysis of additional file data.