# US 2018/0321924 A1 · Classification models for binary code data

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2018/0321924 A1
- Document ID: US20180321924A1
- Status: Published application
- Kind: A1
- Application number: 15/774,819
- Filed: 2015-11-12
- Published: 2018-11-08
- Earliest claimed: 2015-11-12
- Assignees: Entit Software LLC
- Applicants: Entit Software LLC
- Inventors: Amit Bezalel; Amir Wigelman; Roman Gelfand
- Web page: https://patentgrep.com/patent/US20180321924A1

## CPC

- G06F 8/43: {Checking; Contextual analysis}
- G06F 8/35: model driven
- G06F 8/65: Updates (security arrangements therefor G06F21/57)
- G06F 8/654: using techniques specially adapted for alterable solid state memories, e.g. for EEPROM or flash memories
- G06N 20/00: Machine learning
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Examples disclosed herein relate to classification models for binary code data. Some of the examples enable obtaining changed binary code data and unchanged binary code data, and generating, using a machine-leaning algorithm, a classification model based on training data that comprises the changed binary code data and the unchanged binary code data.

## Claims

### Claim 1

A method for generating a classification model for binary code data, the method comprising:
obtaining changed binary code data comprising a first set of binary code data that is produced by a first software code build of a first set of source code, and a second set of binary code data that is produced by a second software code build of a second set of source code;
obtaining unchanged binary data comprising a third set of binary code data that is produced by a third software code build of a third set of source code, and a fourth set of binary code data that is produced by a fourth software code build of the third set of source code; and
generating, using a machine-learning algorithm, a classification model based on training data that comprises the changed binary code data and the unchanged binary code data.

### Claim 2 (depends on 1)

The method of claim 1, wherein the changed binary code data comprises a pair of a first binary code file from the first set of binary code data and a second binary code file from the second set of binary code data, and wherein the unchanged binary code data comprises a pair of a third binary code file from the third set of binary code data and a fourth binary code file from the fourth set of binary code data.

### Claim 3 (depends on 1)

The method of claim 1, further comprising:
obtaining a pair of binary code files to be classified using the classification model; and
determining, using the classification model, whether the pair of binary code files is classified into a first classification group that the changed binary code data belongs to or a second classification group that the unchanged binary code data belongs to.

### Claim 4 (depends on 3)

The method of claim 3, wherein determining that the pair of binary code files is classified into the first classification group further comprises:
determining that one binary code file of the pair includes a portion that is different from the other binary code file of the pair.

### Claim 5 (depends on 4)

The method of claim 4, further comprising:
identifying the portion that is different from the other binary code file of the pair; and
including, in a software patch, the portion that is different from the other binary code file of the pair.

### Claim 6 (depends on 1)

The method of claim 1, wherein the second set of source code includes a portion that is different from the first set of source code.

### Claim 7

A non-transitory machine-readable storage medium comprising instructions executable by a processor of a computing device for generating a classification model for binary code data, the machine-readable storage medium comprising:
instructions to identify a classification model that is generated based on training data comprising: (i) a first pair of binary code files produced by compiling a first pair of source code files that have a portion that is different from each other, and (ii) a second pair of binary code files produced by compiling a second pair of source code files that are identical;
instructions to obtain a third pair of binary code files to be classified using the classification model; and
instructions to determine, using the classification model, whether the third pair of binary code files is classified into a first classification group that includes the first pair of binary code files or a second classification group that includes the second pair of binary code files.

### Claim 8 (depends on 7)

The non-transitory machine-readable storage medium of claim 7, further comprising:
in response to determining that the third pair of binary code files is classified into the first classification group, instructions to identify a portion that is different between the third pair of binary code files; and
instructions to include the identified portion in a software patch.

### Claim 9 (depends on 7)

The non-transitory machine-readable storage medium of claim 7, wherein the instructions to determine using the classification model, whether the third pair of binary code files is classified into the first classification group or the second classification group further comprise:
instructions to compare a first binary code file of the third pair with a second binary code file of the third pair; and
instructions to generate a particular vector to be placed in a vector space of the classification model, the particular vector representing a result of the comparison.

### Claim 10 (depends on 9)

The non-transitory machine-readable storage medium of claim 9, wherein the second binary code file comprises a predetermined number of sections of equal byte size, further comprising:
instructions to identify, for each section of the second binary code file, a portion that is different from the first binary code file;
instructions to determine a byte size of the portion that is different from the first binary code file; and
instructions to determine a percentage of the byte size of the portion in a total byte size of the each section.

### Claim 11 (depends on 9)

The non-transitory machine-readable storage medium of claim 9, wherein the vector space of the classification model comprises the first classification group that a first set of vectors are placed within and the second classification group that a second set of vectors are placed within, further comprising:
in response to determining that the particular vector is placed in the first classification group, instructions to determine that the third pair of binary code files is classified into the first classification group.

### Claim 12

A system for generating a classification model for binary code data comprising:
a processor that:
obtains changed binary code data comprising a first set of binary code data that is produced by a first software code build of a first set of source code, and a second set of binary code data that is produced by a second software code build of a second set of source code;
obtains unchanged binary code data comprising a third set of binary code data that is produced by a third software code build of a third set of source code, and a fourth set of binary code data that is produced by a fourth software code build of the third set of source code;
generates, using a machine-learning algorithm, a classification model based on training data that comprises the changed binary code data and the unchanged binary code data;
classifies a pair of binary files using the classification model; and
determines whether to include a portion of the pair of binary files in a software patch based on the classification.

### Claim 13 (depends on 12)

The system of claim 12, wherein classifying the pair of binary files using the classification model comprises:
determining, using the classification model, whether the pair of binary code files is classified into a first classification group that includes the changed binary code data or a second classification group that includes the unchanged binary code data belongs to.

### Claim 14 (depends on 13)

The system of claim 13, the processor that:
identifies the portion that is different between the pair of binary files; and
in response to determining that the pair of binary code files is classified into the first classification group, includes the portion in the software patch.

### Claim 15 (depends on 13)

The system of claim 13, the processor that:
determines a level of confidence in classifying the pair of binary files into the second classification group;
compares the confidence level with a threshold level; and
in response to determining that the confidence level is below the threshold level, classifying the pair of binary files into the first classification group.
