ML-Based Client Selection for Network Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Complex computer networks face challenges in identifying and diagnosing network anomalies due to the high dimensionality of data and resource constraints on client devices, making it difficult to pinpoint the root cause of issues without overwhelming clients with telemetry data collection.
Innovation Solution
A machine learning-based network assurance service detects anomalies, selects clients exhibiting specific network conditions, initiates targeted network tests, and retrains the anomaly detector based on test results to verify the cause and improve detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If comprehensive telemetry data collection is performed on all clients to diagnose network anomalies, then measurement precision improves, but device complexity and energy consumption increase significantly
Solution Approach 1:
The patent segments the anomaly detection process by dividing clients into different groups based on their characteristics and network conditions. Instead of uniformly collecting data from all clients, the system selectively monitors specific client groups, reducing individual device complexity while maintaining overall detection precision through aggregated insights from multiple segments.
Solution Approach 2:
The patent introduces a network assurance service as an intermediary between clients and the anomaly detection system. This service collects and processes telemetry data centrally, acting as a mediator that aggregates information from multiple clients without requiring complex local processing on each client device, thereby reducing device complexity while preserving measurement precision.
2Measurement precision
If comprehensive telemetry data collection is performed on all clients to diagnose network anomalies, then measurement precision improves, but use of energy increases significantly
Solution Approach 1:
The patent segments the monitoring workload by selectively assigning telemetry collection tasks to specific client groups rather than requiring all clients to continuously collect and transmit data. This selective approach reduces energy consumption for individual devices while maintaining anomaly detection accuracy through strategic sampling from multiple segments.
Solution Approach 2:
The patent applies partial action by collecting telemetry data from a subset of clients rather than all clients continuously. The network assurance service strategically selects which clients to monitor based on network conditions and anomaly patterns, reducing overall energy consumption while maintaining sufficient measurement precision for effective anomaly detection.
3Adaptability or versatility
If machine learning model is continuously retrained with new data, then adaptability improves, but loss of time for model maintenance increases
Solution Approach 1:
The patent implements periodic action by retraining the machine learning model at scheduled intervals rather than continuously. The network assurance service periodically updates the anomaly detection model with newly collected telemetry data, maintaining adaptability to changing network conditions while avoiding the time loss associated with continuous retraining operations.
Data Source
AI summary
In one embodiment, a network assurance service that monitors a network detects a network anomaly in the network using a machine learning-based anomaly detector. The network assurance service identifies a set of network conditions associated with the detected network anomaly. The network assurance service initiates a network test on one or more clients in the network that exhibit the identified network conditions. The network assurance service retrains the machine learning-based anomaly detector based on a result of the network test.