Management entity, network element, system, and methods for supporting anomaly detection for communication networks

A system for supporting anomaly detection for a communication network, the system comprising a management entity and a set of network elements (NEs),

• wherein each NE of the set of NEs is configured to:
• determine, based on a local anomaly detection model, at least one point to be an anomalous point in key performance indicator (KPI) time-series data associated with the communication network;
• provide the anomalous point to the management entity;
• receive, from the management entity, a respective label for each respective point of a set of points of the KPI time series data, wherein the set of points includes the anomalous point provided to the management entity;
• and
• update the local anomaly detection model based on the labels for the set of points;
• wherein the management entity is configured to:
• receive, from each respective NE of the one or more NEs, one or more parameters of the updated local anomaly detection model of the respective NE;
• and
• update a global anomaly detection model based on the one or more received parameters, wherein updating the global anomaly detection model comprises integrating a federated aggregation of the one or more received parameters into a weight matrix of a neural network of the global anomaly detection model.

A method for a management entity for supporting anomaly detection in a communication network, the method comprising:

• receiving, from each of one or more network elements (NEs)) of a set of NEs, at least one point determined to be an anomalous point, wherein the one or more points belong to key performance indicator (KPI) time-series data associated with the communication network;
• providing, to each of the one or more NEs, a label for each point of the one or more points;
• receiving, from each respective NE of the one or more NEs, at least one parameter of an updated local anomaly detection model of the respective NE;
• and
• updating a global anomaly detection model based on the one or more parameters;
• wherein updating the global anomaly detection model comprises integrating a federated aggregation of the one or more received parameters into a weight matrix of a neural network of the global anomaly detection model.

A method for a network element (NE) for supporting anomaly detection for a communication network, the method comprising:

• determining, based on a local anomaly detection model, at least one point to be an anomalous point in key performance indicator (KPI) time-series data associated with the communication network;
• providing the anomalous point to a management entity;
• receiving a label for each point of a set of points of the KPI time series data from the management entity, wherein the set of points includes the anomalous point provided to the management entity;
• updating the local anomaly detection model based on the one or more labels;
• and
• providing at least one parameter of the updated local anomaly detection model to the management entity, wherein the management entity is configured to update a global anomaly detection model by integrating a federated aggregation of one or more received parameters into a weight matrix of a neural network of the global anomaly detection model.