Collaborative Telemetry Engineering for Decentralized Anomaly Detection

The patent segments the centralized anomaly detection function into distributed components at each network device. Each device independently performs deep anomaly detection on its own telemetry data using local machine learning models, eliminating the need to aggregate and process all telemetry data through a single centralized controller. This segmentation reduces computational overhead at any single point while maintaining comprehensive monitoring coverage across the network.

Network devices are empowered to perform self-service anomaly detection by generating and analyzing their own feature vectors locally. Each device uses its own computational resources to run machine learning models and detect anomalies in its telemetry data, rather than relying on external centralized processing. This self-service approach distributes the computational burden and reduces overall system resource consumption.

The system segments the anomaly detection functionality into independent modules deployed at each network device. Each device runs its own machine learning model locally, creating multiple simplified detection units rather than one complex centralized system. This segmentation maintains detection precision through specialized local models while reducing overall system complexity by distributing functionality across standard network devices.

The patent uses copying by deploying identical or similar machine learning models across multiple network devices. Each device maintains a local copy of the anomaly detection model, enabling consistent detection precision across the network without requiring a single complex centralized analysis system. This copying approach simplifies the overall architecture by using replicated simple components rather than a single complex component.