KPI Trajectory-Based Anomaly Detection in Network Assurance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current network assurance systems lack effective methods to detect anomalous behavior in networking devices based on key performance indicators (KPIs), particularly failing to represent and analyze the dynamic changes and correlations between KPIs over time, which are crucial for identifying outliers and anomalies.
Innovation Solution
A network assurance service that utilizes machine learning-based models to represent KPI changes over time as trajectories, applying metrics such as off-directions, off-lengths, and off-speeds to identify anomalous behavior, and provides an indication of such behavior to user interfaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional network assurance methods are used to monitor KPIs, then basic network health tracking is maintained, but the ability to detect anomalous behavior and outliers is insufficient
Solution Approach 1:
The patent transforms KPI data from static scalar values into dynamic trajectory objects that incorporate temporal evolution, directional changes, and relational dynamics. This dimensional transformation enables the detection system to analyze not just KPI values but also their rates of change, patterns of correlation, and deviation from expected trajectories, thereby significantly improving anomaly detection precision without proportionally increasing system complexity
Solution Approach 2:
The system changes the analytical parameters from simple KPI thresholds to trajectory-based metrics including off-directions, off-lengths, and off-speeds. These parameter transformations allow the system to detect anomalies based on the dynamic behavior and relationships between KPIs over time, rather than relying on static threshold comparisons, thus enhancing detection capability while maintaining manageable complexity
2Reliability
If comprehensive KPI monitoring is implemented across the network, then network health assessment capability is improved, but the complexity of analyzing dynamic relationships between multiple KPIs increases
Solution Approach 1:
The patent segments the complex task of multi-KPI analysis into distinct trajectory components (off-directions, off-lengths, off-speeds) that can be independently calculated and analyzed. Each KPI is transformed into a trajectory object with specific attributes, allowing the system to process and evaluate dynamic relationships systematically rather than attempting to analyze all KPI interactions simultaneously, thereby improving assessment reliability while controlling analytical complexity
Solution Approach 2:
The trajectory representation serves as an intermediary data structure between raw KPI measurements and anomaly detection algorithms. By introducing this intermediate layer that captures temporal and relational dynamics, the system can reliably assess network health based on comprehensive KPI monitoring without directly confronting the full complexity of multi-dimensional KPI relationships, as the trajectory abstraction simplifies the analysis burden
3Measurement precision
If static threshold-based anomaly detection is used, then implementation simplicity is maintained, but detection accuracy for dynamic network behaviors is insufficient
Solution Approach 1:
The patent replaces static threshold-based detection with dynamic trajectory-based analysis that adapts to changing network conditions. Instead of fixed thresholds, the system evaluates KPI trajectories against dynamic patterns of normal behavior, incorporating temporal evolution and relational dynamics. This dynamic approach significantly improves detection accuracy for evolving network anomalies while the automated trajectory computation maintains operational simplicity through systematic algorithms
Data Source
AI summary
In one embodiment, a network assurance service that monitors a network receives a plurality of key performance indicators (KPIs) for a networking device in the network over time. The network assurance service represents relationship changes between the KPIs over time as a set of one or more KPI trajectories. The network assurance service uses a machine learning-based model to determine that a behavior of the networking device is anomalous, based on the one or more KPI trajectories. The network assurance service provides an indication of the anomalous behavior of the networking device to a user interface.