# US 2018/0322287 A1 · Machine learning model for malware dynamic analysis

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

- Number: US 2018/0322287 A1
- Document ID: US20180322287A1
- Status: Published application
- Kind: A1
- Application number: 15/588,131
- Filed: 2017-05-05
- Published: 2018-11-08
- Earliest claimed: 2017-05-05
- Assignees: Cylance Inc.
- Applicants: Cylance Inc.
- Inventors: Xuan Zhao; Aditya Kapoor; Matthew Wolff; Andrew Davis; Derek Soeder; Ryan Permeh
- Web page: https://patentgrep.com/patent/US20180322287A1

## CPC

- G06F 21/566: {Dynamic detection, i.e. detection performed at run-time, e.g. emulation, suspicious activities}
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/09: Supervised learning
- G06N 5/025: Extracting rules from data
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 20/00: Machine learning
- H04L 63/145: {the attack involving the propagation of malware through the network, e.g. viruses, trojans or worms}
- G06F 3/048: Interaction techniques based on graphical user interfaces [GUI]
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G06N 20/20: Ensemble learning
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

In some implementations there may be provided a system. The system may include a processor and a memory. The memory may include program code which causes operations when executed by the processor. The operations may include analyzing a series of events contained in received data. The series of events may include events that occur during the execution of a data object. The series of events may be analyzed to at least extract, from the series of events, subsequences of events. A machine learning model may determine a classification for the received data. The machine learning model may classify the received data based at least on whether the subsequences of events are malicious. The classification indicative of whether the received data is malicious may be provided. Related methods and articles of manufacture, including computer program products, are also disclosed.

## Claims

### Claim 1

A system comprising:
at least one processor; and
at least one memory including program code which when executed by the at least one processor causes operations comprising:
analyzing a series of events contained in received data, the series of events comprising events that occur during the execution of a data object, and the series of events being analyzed to at least extract, from the series of events, one or more subsequences of events;
determining, by a machine learning model, a classification for the received data, the machine learning model classifying the received data based at least on whether the one or more subsequences of events are malicious; and
providing the classification indicative of whether the received data is malicious.

### Claim 2 (depends on 1)

The system according to claim 1, wherein the one or more subsequences of events are extracted by at least:
determining a plurality of subsequences in the series of events;
identifying a plurality of most frequent subsequences in the plurality of subsequences; and
selecting, from the plurality of the most frequent subsequences, the one or more subsequences of events.

### Claim 3 (depends on 2)

The system according to claim 2, wherein the selecting of the one or more subsequence of events is based on a frequency of occurrence of the plurality of most frequent subsequences exceeding a predetermined threshold.

### Claim 4 (depends on 2)

The system according to claim 2, wherein the plurality of the most frequent subsequences are identified by at least applying an Apriori algorithm.

### Claim 5 (depends on 4)

The system according to claim 4, wherein the Apriori algorithm is configured to at least:
identify a most frequent event; and
identify, based at least on the most frequent event, a most frequent subsequence of events, the most frequent subsequence of events including the most frequent event and at least one other event.

### Claim 6 (depends on 1)

The system according to claim 1, wherein the machine learning model comprises a recurrent neural network.

### Claim 7 (depends on 6)

The system according to claim 6, wherein the recurrent neural network comprises a long short term memory network.

### Claim 8 (depends on 7)

The system according to claim 7, wherein the long short term memory network comprises a memory cell, the memory cell being configured maintain a value, the memory cell being further configured to determine, based at least on an event, whether to update, output, discard, and/or continue to maintain the value.

### Claim 9 (depends on 1)

The system according to claim 1, wherein the machine learning model determines the classification for the received data by at least:
generating a representation the one or more subsequences of events;
determining an average representation of the one or more subsequences of events over a plurality of time steps associated with the series of events;
determining a logistic regression of an average representation of the one or more subsequences of events; and
determining, based on the logistic regression, the classification for the received data.

### Claim 10 (depends on 1)

The system according to claim 1, further comprising:
determining whether the series of events includes at least one malicious event; and
classifying the received data based at least on a presence of the at least one malicious event in the series of events.

### Claim 11 (depends on 10)

The system according to claim 10, further comprising:
determining that an event is malicious based at least on a filename and/or a filepath associated with the event.

### Claim 12 (depends on 11)

The system according to claim 11, wherein the filename and/or the filepath associated with the event is analyzed based at least on a pattern associated with one or more filenames and/or filepaths that are known to be non-malicious.

### Claim 13 (depends on 12)

The system according to claim 12, wherein the pattern associated with the one or more non-malicious filenames and/or filepaths is modeled using a Markov chain.

### Claim 14 (depends on 11)

The system according to claim 11, wherein the data object comprises a file, a function, and/or a software program.

### Claim 15 (depends on 1)

The system according to claim 1, wherein the one or more subsequence of events comprise consecutive and/or non-consecutive events from the series of events.

### Claim 16

A computer-implemented method, comprising:
analyzing a series of events contained in received data, the series of events comprising events that occur during the execution of a data object, and the series of events being analyzed to at least extract, from the series of events, one or more subsequences of events;
determining, by a machine learning model, a classification for the received data, the machine learning model classifying the received data based at least on whether the one or more subsequences of events are malicious; and
providing the classification indicative of whether the received data is malicious.

### Claim 17 (depends on 16)

The computer-implemented method according to claim 16, wherein the one or more subsequences of events are extracted by at least:
determining a plurality of subsequences in the series of events;
identifying a plurality of most frequent subsequences in the plurality of subsequences; and
selecting, from the plurality of the most frequent subsequences, the one or more subsequences of events.

### Claim 18 (depends on 17)

The computer-implemented method according to claim 17, wherein the selecting of the one or more subsequence of events is based on a frequency of occurrence of the plurality of most frequent subsequences exceeding a predetermined threshold.

### Claim 19 (depends on 17)

The computer-implemented method according to claim 17, wherein the plurality of the most frequent subsequences are identified by at least applying an Apriori algorithm.

### Claim 20 (depends on 19)

The computer-implemented method according to claim 19, wherein the Apriori algorithm is configured to at least:
identify a most frequent event; and
identify, based at least on the most frequent event, a most frequent subsequence of events, the most frequent subsequence of events including the most frequent event and at least one other event.

### Claim 21 (depends on 16)

The computer-implemented method according to claim 16, wherein the machine learning model comprises a recurrent neural network.

### Claim 22 (depends on 21)

The computer-implemented method according to claim 21, wherein the recurrent neural network comprises a long short term memory network.

### Claim 23 (depends on 22)

The computer-implemented method according to claim 22, wherein the long short term memory network comprises a memory cell, the memory cell being configured maintain a value, the memory cell being further configured to determine, based at least on an event, whether to update, output, discard, and/or continue to maintain the value.

### Claim 24 (depends on 16)

The computer-implemented method according to claim 16, wherein the machine learning model determines the classification for the received data by at least:
generating a representation the one or more subsequences of events;
determining an average representation of the one or more subsequences of events over a plurality of time steps associated with the series of events;
determining a logistic regression of an average representation of the one or more subsequences of events; and
determining, based on the logistic regression, the classification for the received data.

### Claim 25 (depends on 16)

The computer-implemented method according to claim 16, further comprising:
determining whether the series of events includes at least one malicious event; and
classifying the received data based at least on a presence of the at least one malicious event in the series of events.

### Claim 26 (depends on 25)

The computer-implemented method according to claim 25, further comprising:
determining that an event is malicious based at least on a filename and/or a filepath associated with the event.

### Claim 27 (depends on 26)

The computer-implemented method according to claim 26, wherein the filename and/or the filepath associated with the event is analyzed based at least on a pattern associated with one or more filenames and/or filepaths that are known to be non-malicious.

### Claim 28 (depends on 27)

The computer-implemented method according to claim 27, wherein the pattern associated with the one or more non-malicious filenames and/or filepaths is modeled using a Markov chain.

### Claim 29 (depends on 26)

The computer-implemented method of claim 26, wherein the data object comprises a file, a function, and/or a software program.

### Claim 30 (depends on 16)

The computer-implemented method according to claim 16, wherein the one or more subsequence of events comprise consecutive and/or non-consecutive events from the series of events.

### Claim 31

A non-transitory computer-readable storage medium including program code, which when executed by at least one data processor, cause operations comprising:
analyzing a series of events contained in received data, the series of events comprising events that occur during the execution of a data object, and the series of events being analyzed to at least extract, from the series of events, one or more subsequences of events;
determining, by a machine learning model, a classification for the received data, the machine learning model classifying the received data based at least on whether the one or more subsequences of events are malicious; and
providing the classification indicative of whether the received data is malicious.

### Claim 32

An apparatus, comprising:
means for analyzing a series of events contained in received data, the series of events comprising events that occur during the execution of a data object, and the series of events being analyzed to at least extract, from the series of events, one or more subsequences of events;
means for determining, by a machine learning model, a classification for the received data, the machine learning model classifying the received data based at least on whether the one or more subsequences of events are malicious; and
means for providing the classification indicative of whether the received is malicious.
