# US 2025/0168181 A1 · Multi-layer anomaly detector

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

- Number: US 2025/0168181 A1
- Document ID: US20250168181A1
- Status: Published application
- Kind: A1
- Application number: 18/512,867
- Filed: 2023-11-17
- Published: 2025-05-22
- Earliest claimed: 2023-11-17
- Assignees: Microsoft Technology Licensing, LLC
- Applicants: Microsoft Technology Licensing, LLC
- Inventors: Zhongkai LIU
- Web page: https://patentgrep.com/patent/US20250168181A1

## CPC

- G06F 21/566: {Dynamic detection, i.e. detection performed at run-time, e.g. emulation, suspicious activities}
- H04L 63/1425: {Traffic logging, e.g. anomaly detection}
- G06F 21/55: Detecting local intrusion or implementing counter-measures
- G06F 21/554: {involving event detection and direct action}
- G06N 3/045: Combinations of networks
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 20/20: Ensemble learning
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- 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

Multi-layer anomaly detection identifies and issues alerts to focus limited resources on the most concerning activities detected in voluminous data. A multi-layer anomaly detector includes a clusterer, a forecaster, and a statistics generator to respectively generate a first, second, and third anomaly lists for input data indicating voluminous events relative to one or more domains, such as access to computing devices; access to real estate; or financial transactions. An ensemble detector may generate an ensemble anomaly list indicating a subset of the voluminous events based on the first, second, and third anomaly lists. The lists may indicate anomaly (e.g., security risk) scores. The ensemble anomaly list may combine the anomaly (e.g., risk) scores in the first, second, and third anomaly lists. An identifier may generate an alert for the subset of events, e.g., with relative security risk scores for the relevant individuals, entities, and/or events.

## Claims

### Claim 1

A multi-layer anomaly detector system comprising:
a processor; and
a memory device that stores program code configured to be executed by the processor, the program code comprising:
a first type anomaly detector configured to generate a first anomaly list for input data that indicates voluminous events;
a second type anomaly detector configured to generate a second anomaly list for the input data;
a third type anomaly detector configured to generate a third anomaly list for the input data, the first, the second, and the third type anomaly detectors being different anomaly detector types;
an ensemble detector configured to generate an ensemble anomaly list based on the first, second, and third anomaly lists; and
an alerter configured to generate one or more alerts for a subset of events indicated by the ensemble anomaly list.

### Claim 2 (depends on 1)

The multi-layer anomaly detector system of claim 1, wherein the input data indicates behavior of at least one of individuals or entities.

### Claim 3 (depends on 2)

The multi-layer anomaly detector system of claim 2, wherein the input data indicates at least one of the following behaviors of the at least one of individuals or entities:
accesses to computing devices;
accesses to real estate; or
financial transactions.

### Claim 4 (depends on 1)

The multi-layer anomaly detector system of claim 1,
wherein the first type anomaly detector comprises a clusterer;
wherein the second type anomaly detector comprises a forecaster; and
wherein the third type anomaly detector comprises a statistics generator.

### Claim 5 (depends on 4)

The multi-layer anomaly detector system of claim 4, wherein the clusterer comprises a self-organizing map (SOM) neural network model.

### Claim 6 (depends on 4)

The multi-layer anomaly detector system of claim 4, wherein the forecaster is configured to select a forecasting model from a plurality of forecasting models to generate forecast data for individuals or entities indicated in the input data.

### Claim 7 (depends on 4)

The multi-layer anomaly detector system of claim 4, wherein the statistics generator is configured to generate domain-specific statistics.

### Claim 8 (depends on 4)

The multi-layer anomaly detector system of claim 4, wherein the first, second, third, and ensemble anomaly lists indicate at least one of anomaly scores or risk scores.

### Claim 9 (depends on 8)

The multi-layer anomaly detector system of claim 8, wherein the ensemble anomaly list combines the at least one of anomaly scores or risk scores of the first, second, and third anomaly lists.

### Claim 10 (depends on 1)

The multi-layer anomaly detector system of claim 1, wherein the multi-layer anomaly detector system comprises anomaly detection as a service for a plurality of domains.

### Claim 11

A computer-implemented method of anomaly detection comprising:
generating, by a first type anomaly detector, a first anomaly list for input data;
generating, by a second type anomaly detector, a second anomaly list for the input data;
generating, by a third type anomaly detector, a third anomaly list for the input data, the first, the second, and the third type anomaly detectors being different anomaly detector types;
generating, by an ensemble anomaly detector, an ensemble anomaly list based on the first, second, and third anomaly lists; and
performing an action with respect to an anomaly indicated in the ensemble anomaly list.

### Claim 12 (depends on 11)

The computer-implemented method of anomaly detection of claim 11, wherein the input data indicates behavior of at least one of individuals or entities.

### Claim 13 (depends on 12)

The computer-implemented method of anomaly detection of claim 12, wherein the input data indicates at least one of the following behaviors of the at least one of individuals or entities:
accesses to computing devices;
accesses to real estate; or
at least one of requested financial transactions or performed financial transactions.

### Claim 14 (depends on 11)

The computer-implemented method of anomaly detection of claim 11,
wherein the first type anomaly detector comprises a clusterer;
wherein the second type anomaly detector comprises a forecaster; and
wherein the third type anomaly detector comprises a statistics generator.

### Claim 15 (depends on 14)

The computer-implemented method of anomaly detection of claim 14, further comprising:
selecting, by the clusterer, a cluster model from a plurality of cluster models to generate cluster data for at least one of individuals or entities indicated in the input data.

### Claim 16 (depends on 14)

The computer-implemented method of anomaly detection of claim 14, further comprising:
selecting, by the forecaster, a forecasting model from a plurality of forecasting models to generate forecast data for at least one of individuals or entities indicated in the input data.

### Claim 17 (depends on 14)

The computer-implemented method of anomaly detection of claim 14, further comprising:
selecting, by the statistics generator, a domain from a plurality of domains to generate domain-specific statistics for at least one of individuals or entities indicated in the input data.

### Claim 18 (depends on 14)

The computer-implemented method of anomaly detection of claim 14, wherein the first, second, third, and ensemble anomaly lists indicate at least one of anomaly scores or risk scores.

### Claim 19 (depends on 18)

The computer-implemented method of anomaly detection of claim 18, wherein the ensemble anomaly list combines the at least one of anomaly scores or risk scores of the first, second, and third anomaly lists.

### Claim 20 (depends on 11)

The computer-implemented method of anomaly detection of claim 11, wherein the multi-layer anomaly detector system comprises anomaly detection as a service for a plurality of domains.
