# US 2023/0412624 A1 · Estimation device, estimation method, and estimation program

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

- Number: US 2023/0412624 A1
- Document ID: US20230412624A1
- Status: Published application
- Kind: A1
- Application number: 18/035,109
- Filed: 2020-11-19
- Published: 2023-12-21
- Earliest claimed: 2020-11-19
- Assignees: Nippon Telegraph and Telephone Corporation
- Applicants: Nippon Telegraph and Telephone Corporation
- Inventors: Tomohiro NAGAI; Masanori YAMADA; Tomokatsu TAKAHASHI; Yasuhiro TERAMOTO; Yuki YAMANAKA
- Web page: https://patentgrep.com/patent/US20230412624A1

## CPC

- H04L 63/1425: {Traffic logging, e.g. anomaly detection}
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 20/00: Machine learning
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

An estimation device includes processing circuitry configured to generate a feature quantity by performing invertible transformation on a payload of a packet character by character with respect to each packet determined to be abnormal or normal by an abnormality detector, and give a determination result as to whether a packet is abnormal or normal to the generated feature quantity, learn a model that classifies whether the packet is abnormal or normal by machine learning using the feature quantity of the payload of the packet and the determination result as to whether the packet is abnormal or normal as teacher data, extract a number of dimensions of the feature quantity in which a contribution degree to classification is equal to or greater than a predetermined value in the learned model, and estimate a cause part of abnormality in the payload of the packet determined to be abnormal.

## Claims

### Claim 1

An estimation device comprising:
processing circuitry configured to:
generate a feature quantity by performing invertible transformation on a payload of a packet character by character with respect to each packet determined to be abnormal or normal by an abnormality detector, and give a determination result as to whether a packet is abnormal or normal to the generated feature quantity;
learn a model that classifies whether the packet is abnormal or normal by machine learning using the feature quantity of the payload of the packet and the determination result as to whether the packet is abnormal or normal as teacher data;
extract a number of dimensions of the feature quantity in which a contribution degree to classification is equal to or greater than a predetermined value in the learned model; and
estimate a cause part of abnormality in the payload of the packet determined to be abnormal using the extracted number of dimensions of the feature quantity, and output a result of estimation.

### Claim 2 (depends on 1)

The estimation device according to claim 1, wherein
the model is
a model using a decision tree, and
the processing circuitry is further configured to:
extract the number of dimensions of the feature quantity written in a branch condition from a node in which the branch condition in the decision tree obtained by the machine learning is written as the number of dimensions of the feature quantity in which the contribution degree is equal to or greater than a predetermined value.

### Claim 3 (depends on 1)

The estimation device according to claim 1, wherein
the model is
a model using linear regression or logistic regression.

### Claim 4 (depends on 1)

The estimation device according to claim 1, wherein
the processing circuitry is further configured to:
specify a part estimated as a cause part of abnormality in the payload of the packet determined to be abnormal based on the extracted number of dimensions of the feature quantity, and outputs information obtained by visualizing the specified part as a result of the estimation.

### Claim 5 (depends on 1)

The estimation device according to claim 1, wherein
the invertible transformation
transforms a character string into a numeric string according to ASCII code table.

### Claim 6

An estimation method comprising:
generating a feature quantity by performing invertible transformation on a payload of a packet character by character with respect to each packet determined to be abnormal or normal by an abnormality detector, and giving a determination result as to whether a packet is abnormal or normal to the generated feature quantity;
learning a model that classifies whether the packet is abnormal or normal by machine learning using the feature quantity of the payload of the packet and the determination result as to whether the packet is abnormal or normal as teacher data;
extracting a number of dimensions of the feature quantity in which a contribution degree to classification is equal to or greater than a predetermined value in the learned model; and
estimating a cause part of abnormality in the payload of the packet determined to be abnormal using the extracted number of dimensions of the feature quantity, and outputting a result of estimation.

### Claim 7

A non-transitory computer-readable recording medium storing therein an estimation program that causes a computer to execute a process comprising:
generating a feature quantity by performing invertible transformation on a payload of a packet character by character with respect to each packet determined to be abnormal or normal by an abnormality detector, and giving a determination result as to whether a packet is abnormal or normal to the generated feature quantity;
learning a model that classifies whether the packet is abnormal or normal by machine learning using the feature quantity of the payload of the packet and the determination result as to whether the packet is abnormal or normal as teacher data;
extracting a number of dimensions of the feature quantity in which a contribution degree to classification is equal to or greater than a predetermined value in the learned model; and
estimating a cause part of abnormality in the payload of the packet determined to be abnormal using the extracted number of dimensions of the feature quantity, and outputting a result of estimation.
