Granted Patent B2

US 12,430,429 · App. 17/572,548 · Granted Sep 30, 2025

Inventors: Benjamin Chang (Fremont, CA); Ghanashyam Satpathy (Bangalore, IN)

Assignee: Netskope, Inc.

G06F21/56G06F21/53G06N5/01G06N20/00

Abstract

Claims (28)

1. A method for classifying input documents in a networked system to determine if at least one of the documents may include malicious code, comprising the steps of:

parsing a document file to separate code data from document payload data;

generating for the document file at least obfuscation features indicative of past instances of malware embedded in code using known obfuscation methods;

inputting the obfuscation features to a trained machine learning model and applying the trained machine learning model to process the document file to predict the presence of malicious code and classify the document file as to its threat level;

using a secondary malware detection engine operating to reduce false positives;

classifying the document file, based on the threat level predicted by the trained machine learning model as qualified by the secondary malware detection engine, as safe, suspicious, or malicious; and

based on the step of classifying, accepting the document file classified as safe into the networked system, or blocking the document file classified as malicious, or isolating the document file classified as suspicious for threat analysis.

2. The method of claim 1 wherein a network interface that is coupled to a network, and where receiving the document file occurs via said network interface.

3. The method of claim 1 wherein said document file is a Microsoft Office document.

4. The method of claim 1 wherein said document file is one of a Microsoft Word document, a Microsoft Excel document, or a Microsoft Power Point document.

5. The method of claim 1 wherein said document file is one of a word processing document, a spreadsheet document or a presentation document.

6. The method of claim 1 wherein the code includes a VBA macro.

7. The method of claim 1 wherein said obfuscation scoring features include macro related features, said features including at least one use of CreateObject, Shell, FileSystem, URLDownloadToFile, CallByName, or Detect Sandbox.

8. The method of claim 1 further including testing for malicious code within a document file while quarantining said document file into a sandbox.

9. A system including one or more processors coupled to memory, the memory loaded with computer instructions to classify input documents in a networked system to determine if at least one of the documents may include malicious code, the instructions, when executed on the processors, implement actions comprising:

parsing a document file to separate code data from document payload data;

generating for the document file at least obfuscation features indicative of past instances of malware embedded in code using known obfuscation methods;

inputting the obfuscation features to a trained machine learning model and applying the trained machine learning model to process the document file to predict the presence of malicious code and classify the document file as to its threat level;

using a secondary malware detection engine operating to reduce false positives;

classifying the document file, based on the threat level predicted by the trained machine learning model as qualified by the secondary malware detection engine, as safe, suspicious, or malicious; and

based on the step of classifying, accepting the document file classified as safe into the networked system, or blocking the document file classified as malicious, or isolating the document file classified as suspicious for threat analysis.

10. The system of claim 9 wherein said system further includes a network interface that is coupled to a network, and where receiving the document file occurs via said network interface.

11. The system of claim 9 wherein said document file is a Microsoft Office document.

12. The system of claim 9 wherein said document file is one of a Microsoft Word document, a Microsoft Excel document, or a Microsoft Power Point document.

13. The system of claim 9 wherein said document file is one of a word processing document, a spreadsheet document or a presentation document.

14. The system of claim 9 wherein the code includes a VBA macro.

15. The system of claim 9 wherein said obfuscation scoring features include macro related features, said features including at least one use of CreateObject, Shell, FileSystem, URLDownloadToFile, CallByName, or Detect Sandbox.

16. The system of claim 9 further including testing for malicious code within a document file while quarantining said document file into a sandbox.

Assignments (1)

Continuity (2)

References Cited (178)