# US 2022/0286263 A1 · Signaling analysis method and related apparatus

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

- Number: US 2022/0286263 A1
- Document ID: US20220286263A1
- Status: Published application
- Kind: A1
- Application number: 17/752,848
- Filed: 2022-05-24
- Published: 2022-09-08
- Earliest claimed: 2019-11-25
- Assignees: Huawei Technologies Co., Ltd.
- Applicants: Huawei Technologies Co., Ltd.
- Inventors: Zhen Qin; Siwei Rao; Qiang Ye; Jia Lv; Guangjian Tian
- Web page: https://patentgrep.com/patent/US20220286263A1

## CPC

- G06N 3/088: Non-supervised learning, e.g. competitive learning
- 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 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/09: Supervised learning
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- H04L 41/344: Out-of-band transfers
- 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

Signaling analysis methods and apparatus are provided to obtain a to-be-detected signaling flow that includes N pieces of signaling. First feature construction is performed on a message type and an information element that are included in each of the N pieces of signaling to obtain a first feature sequence. The first feature sequence includes N first feature vectors which are in a one-to-one correspondence with the N pieces of signaling. The first feature sequence is input to a first signaling anomaly detection model to perform anomaly detection processing. A first anomaly detection result is produced indicating that the to-be-detected signaling flow is normal or abnormal. In this way, an anomaly caused due to an information element error can be covered. The method and the apparatus can be multiplexed between different protocols.

## Claims

### Claim 1

A signaling analysis method, comprising:
obtaining, by a signaling analysis apparatus, a to-be-detected signaling flow, wherein the to-be-detected signaling flow comprises N pieces of signaling, where N is an integer greater than 1;
performing, by a signaling analysis apparatus, first feature construction on a message type and on an information element that are included in each of the N pieces of signaling to obtain a first feature sequence, the first feature sequence comprising N first feature vectors, the N first feature vectors being in a one-to-one correspondence with the N pieces of signaling;
inputting the first feature sequence to a first signaling anomaly detection model;
performing anomaly detection processing; and
outputting a first anomaly detection result indicating that the to-be-detected signaling flow is normal or abnormal.

### Claim 2 (depends on 1)

The method according to claim 1, further comprising:
performing second feature construction on the message type included in each of the N pieces of signaling to obtain a second feature sequence, the second feature sequence comprising N second feature vectors in a one-to-one correspondence with the N pieces of signaling;
inputting the second feature sequence to a second signaling anomaly detection model;
performing anomaly detection processing;
obtaining a second anomaly detection result indicating that the to-be-detected signaling flow is normal or abnormal;
the performing first feature construction comprises:
when the second anomaly detection result indicates that the to-be-detected signaling flow is normal, separately performing the first feature construction on each of the N pieces of signaling to obtain the first feature sequence.

### Claim 3 (depends on 1)

The method according to claim 1, wherein the separately performing first feature construction on a message type and on an information element comprises:
performing feature construction on a combination of a first message type and a target information element to obtain a first vector, wherein the first message type is a message type included in first signaling, the target information element is an information element included in the first signaling, the first signaling is any of the N pieces of signaling, and the first vector is included in the first feature sequence.

### Claim 4 (depends on 3)

The method according to claim 3, wherein the sequentially performing the first feature construction on the N pieces of signaling comprises:
using M information elements in second signaling as text that comprises one or more words in a natural language processing NLP algorithm performing feature construction on the text to obtain a second vector, the second signaling being any of the N pieces of signaling, the second vector being included in the first feature sequence, and M is an integer greater than 1.

### Claim 5 (depends on 2)

The method according to claim 2, wherein separately performing second feature construction on the message type included in each of the N pieces of signaling comprises:
including a second message type as a word in the natural language processing NLP algorithm and performing feature construction on the word to obtain a third vector, the second message type being a message type included in third signaling, the third signaling being any of the N pieces of signaling and the third vector being included in the second feature sequence.

### Claim 6 (depends on 1)

The method according to claim 1, wherein an Fth first feature vector in the first feature sequence corresponds to an Fth piece of signaling in the N pieces of signaling and the inputting the first feature sequence to a first signaling anomaly detection model to perform anomaly detection processing and outputting a first anomaly detection result comprises:
in an Fth round of anomaly detection processing, inputting a third feature sequence to the first signaling anomaly detection model and performing anomaly detection processing to obtain a first set comprising at least one combination of a message type and an information element, feature vectors in the third feature sequence being successively an (F−K)th first feature vector to an (F−1)th first feature vector in the first feature sequence, F being an integer greater than 1, and K being an integer greater than 1 and less than F; and
outputting the first anomaly detection result when a combination of a message type and an information element of the Fth piece of signaling in the N pieces of signaling is not included in the first set, the first anomaly detection result indicating that the to-be-detected signaling flow is abnormal.

### Claim 7 (depends on 6)

The method according to claim 6, further comprising:
performing an (F+1)th round of anomaly detection processing when the combination of the message type and the information element of the Fth piece of signaling is included in the first set and F is less than N; or
outputting the first anomaly detection result when the combination of the message type and the information element of the Fth piece of signaling is included in the first set and F is equal to N, the first anomaly detection result indicating that the to-be-detected signaling flow is normal.

### Claim 8 (depends on 1)

The method according to claim 1, further comprising:
determining an anomaly location in the to-be-detected signaling flow when the first anomaly detection result indicates that the to-be-detected signaling flow is abnormal.

### Claim 9 (depends on 8)

The method according to claim 8, wherein an Hth first feature vector in the first feature sequence corresponds to an Hth piece of signaling in the N pieces of signaling and the determining an anomaly location in the to-be-detected signaling flow comprises:
in an Hth round of signaling anomaly locating, inputting a fifth feature sequence to the first signaling anomaly detection model and performing anomaly detection processing to obtain a third set comprising at least one combination of a message type and an information element, feature vectors in the fifth feature sequence being successively an (H−L)th first feature vector to an (H−1)th first feature vector in the first feature sequence, H being an integer greater than 1 and L is being an integer greater than 1 and less than H; and
when a combination of a message type and an information element of the Hth piece of signaling in the N pieces of signaling is not comprised in the third set, determining that the Hth piece of signaling is abnormal.

### Claim 10 (depends on 8)

The method according to claim 8, wherein an Hth first feature vector in the first feature sequence corresponds to an Hth piece of signaling in the N pieces of signaling, and the determining an anomaly location in the to-be-detected signaling flow comprises:
obtaining an anomaly probability sequence corresponding to the N pieces of signaling, wherein a Gth probability in the anomaly probability sequence represents a probability that first (D+G) pieces of signaling in the N pieces of signaling comprise abnormal signaling, and G and D are both integers greater than 0; and
determining the anomaly location in the to-be-detected signaling flow based on the anomaly probability sequence.

### Claim 11

A signaling analysis apparatus, comprising a processor and a memory, the memory being configured to store program instructions that, when executed by the processor, cause the signaling analysis apparatus to perform operations comprising:
obtaining a to-be-detected signaling flow, wherein the to-be-detected signaling flow includes N pieces of signaling, N being an integer greater than 1;
performing first feature construction on a message type and an information element that are included in each of the N pieces of signaling to obtain a first feature sequence, the first feature sequence comprising N first feature vectors, the N first feature vectors being in a one-to-one correspondence with the N pieces of signaling; and
inputting the first feature sequence to a first signaling anomaly detection model to perform anomaly detection processing to produce a first anomaly detection result indicating that the to-be-detected signaling flow is normal or abnormal.

### Claim 12 (depends on 11)

The apparatus according to claim 11, the program instructions being configured, upon execution by the processor, to cause the signaling analysis apparatus to:
perform operations comprising second feature construction on the message type included in each of the N pieces of signaling to obtain a second feature sequence, the second feature sequence comprising N second feature vectors, the N second feature vectors being in a one-to-one correspondence with the N pieces of signaling; and
input the second feature sequence to a second signaling anomaly detection model to perform anomaly detection processing to produce a second anomaly detection result indicating that the to-be-detected signaling flow is normal or abnormal, the processor being configured to, when the second anomaly detection result indicates that the to-be-detected signaling flow is normal, perform the first feature construction on each of the N pieces of signaling to obtain the first feature sequence.

### Claim 13 (depends on 12)

The apparatus according to claim 12, the program instructions being configured, upon execution by the processor, to cause the signaling analysis apparatus to:
perform, using a second message type as a word in the natural language processing NLP algorithm, feature construction on the word to obtain a third vector, wherein the second message type is a message type included in third signaling, the third signaling being any of the N pieces of signaling, and the third vector being included in the second feature sequence.

### Claim 14 (depends on 11)

The apparatus according to claim 11, the program instructions being configured, upon execution by the processor, to cause the signaling analysis apparatus to:
perform, using a combination of a first message type and a target information element feature construction on the combination to obtain a first vector, wherein the first message type is a message type included in first signaling, the target information element is an information element included in the first signaling, the first signaling is any of the N pieces of signaling, and the first vector is included in the first feature sequence.

### Claim 15 (depends on 14)

The apparatus according to claim 14, the program instructions being configured, upon execution by the processor, to cause the signaling analysis apparatus to:
perform, using M information elements in second signaling as text including one or more words in a natural language processing NLP algorithm, feature construction on the text to obtain a second vector, wherein the second signaling is any of the N pieces of signaling, the second vector includes the first feature sequence, and M is an integer greater than 1.

### Claim 16 (depends on 11)

The apparatus according to claim 11, the program instructions being configured, upon execution by the processor, to cause the signaling analysis apparatus to:
perform anomaly detection processing in an Fth round of anomaly detection processing comprising inputting a third feature sequence to the first signaling anomaly detection model to obtain a first set, wherein the first set comprises at least one combination of a message type and an information element, and feature vectors in the third feature sequence are successively an (F−K)th first feature vector to an (F−1)th first feature vector in the first feature sequence, F being an integer greater than 1 and K being an integer greater than 1 and less than F; and
output the first anomaly detection result when a combination of a message type and an information element of an Fth piece of signaling in the N pieces of signaling is not included in the first set, the first anomaly detection result indicating that the to-be-detected signaling flow is abnormal.

### Claim 17 (depends on 16)

The apparatus according to claim 16, the program instructions being configured, upon execution by the processor, to cause the signaling analysis apparatus to:
perform an (F+1)th round of anomaly detection processing when the combination of the message type and the information element of the Fth piece of signaling is included in the first set and F is less than N; or
output the first anomaly detection result when the combination of the message type and the information element of the Fth piece of signaling is included in the first set and F is equal to N, the first anomaly detection result indicating that the to-be-detected signaling flow is normal.

### Claim 18 (depends on 11)

The apparatus according to claim 11, wherein:
an Fth second feature vector in the second feature sequence corresponds to an Fth piece of signaling in the N pieces of signaling; and
the program instructions are configured, upon execution by the processor, to cause the signaling analysis apparatus to:
input a fourth feature sequence to the second signaling anomaly detection model in an Fth round of anomaly detection processing and perform anomaly detection processing to obtain a second set, wherein the second set comprises at least one message type, and feature vectors in the fourth feature sequence are successively an (F−K)th second feature vector to an (F−1)th second feature vector in the second feature sequence, F being an integer greater than 1 and K being an integer greater than 1 and less than F; and
obtain the second anomaly detection result when a message type of the Fth piece of signaling in the N pieces of signaling is not included in the second set, wherein the second anomaly detection result indicates that the to-be-detected signaling flow is abnormal.

### Claim 19 (depends on 18)

The apparatus according to claim 18, the program instructions being configured, upon execution by the processor, to cause the signaling analysis apparatus to:
perform an (F+1)th round of anomaly detection processing when the message type of the Fth piece of signaling is included in the second set and F is less than N; or
obtain the second anomaly detection result when the message type of the Fth piece of signaling is included in the second set and F is equal to N, wherein the second anomaly detection result indicates that the to-be-detected signaling flow is normal.

### Claim 20 (depends on 11)

The apparatus according to claim 11, the program instructions being configured, upon execution by the processor, to cause the signaling analysis apparatus to determine an anomaly location in the to-be-detected signaling flow when the first anomaly detection result indicates that the to-be-detected signaling flow is abnormal.

### Claim 21 (depends on 20)

The apparatus according to claim 20, wherein an Hth first feature vector in the first feature sequence corresponds to an Hth piece of signaling in the N pieces of signaling; and
the program instructions are configured, upon execution by the processor, to cause the signaling analysis apparatus to:
input a fifth feature sequence to the first signaling anomaly detection model in an Hth round of signaling anomaly locating, and perform anomaly detection processing to obtain a third set, wherein the third set comprises at least one combination of a message type and an information element, and feature vectors in the fifth feature sequence are successively an (H−L)th first feature vector to an (H−1)th first feature vector in the first feature sequence, H is an integer greater than 1, and L is an integer greater than 1 and less than H; and
determine that the Hth piece of signaling is abnormal when a combination of a message type and an information element of the Hth piece of signaling in the N pieces of signaling is not included in the third set.

### Claim 22 (depends on 20)

The apparatus according to claim 20, wherein an Hth first feature vector in the first feature sequence corresponds to an Hth piece of signaling in the N pieces of signaling; and
the program instructions are configured, upon execution by the processor, to cause the signaling analysis apparatus to:
obtain an anomaly probability sequence corresponding to the N pieces of signaling, wherein a Gth probability in the anomaly probability sequence represents a probability that first (D+G) pieces of signaling in the N pieces of signaling comprise abnormal signaling, and G and D are both integers greater than 0; and
determine the anomaly location in the to-be-detected signaling flow based on the anomaly probability sequence.

### Claim 23 (depends on 22)

The apparatus according to claim 22, the program instructions being configured, upon execution by the processor, to cause the signaling analysis apparatus to:
input a sixth feature sequence to the first signaling anomaly detection model, and perform anomaly detection processing to obtain a Gth probability in the anomaly probability sequence, wherein feature vectors comprised in the sixth feature sequence are successively first (D+G) first feature vectors in the first feature sequence.

### Claim 24 (depends on 22)

The apparatus according to claim 22, the program instructions being configured, upon execution by the processor, to cause the signaling analysis apparatus is specifically configured to:
determine that signaling in a first signaling interval is abnormal when a difference between the Gth probability and a (G−1)th probability in the anomaly probability sequence is greater than a probability threshold, wherein the first signaling interval comprises a (G+D−1)th piece of signaling to an Nth piece of signaling in the N pieces of signaling and G is an integer greater than 1;
determine that signaling in a second signaling interval is abnormal when each probability in the anomaly probability sequence is not less than a previous probability, wherein the second signaling interval comprises a (P+D)th piece of signaling to the Nth piece of signaling in the N pieces of signaling, a difference between a Pth probability and a (P+1)th probability in the anomaly probability sequence is not less than a difference between any two adjacent probabilities in the anomaly probability sequence, and P is an integer greater than 0;
determine that signaling in a third signaling interval is abnormal when a probability in the anomaly probability sequence increases from a first value to a second value and decreases from a third value to the first value before the second value is maintained, wherein the first value is less than a first threshold, the second value and the third value are both greater than a second threshold, the first threshold is less than the second threshold, the third signaling interval comprises a (Q+D)th piece of signaling to the Nth piece of signaling in the N pieces of signaling, a Qth probability in the anomaly probability sequence is a probability at a start point of a final rise section in an anomaly probability curve, and Q is an integer greater than 0; and
determine that signaling in a fourth signaling interval is abnormal when each probability in the anomaly probability sequence is not less than the probability threshold, wherein the fourth signaling interval comprises a Dth piece of signaling to the Nth piece of signaling in the N pieces of signaling.

### Claim 25

A computer-readable storage medium comprising a non-transitory computer-readable storage medium storing a computer program including program instructions to perform:
obtaining a to-be-detected signaling flow, wherein the to-be-detected signaling flow comprises N pieces of signaling, and N is an integer greater than 1;
perform first feature construction on a message type and an information element that are included in each of the N pieces of signaling to obtain a first feature sequence, the first feature sequence comprising N first feature vectors, the N first feature vectors being in a one-to-one correspondence with the N pieces of signaling;
inputting the first feature sequence to a first signaling anomaly detection model to perform anomaly detection processing; and
outputting a first anomaly detection result indicates that the to-be-detected signaling flow is normal or abnormal.
