Network Anomaly Detection via Dimensionality-Reduced KPI Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increased complexity and volume of data generated by virtualized and containerized 5G network functions (VNFs/CNFs) pose challenges in detecting anomalous behavior due to proprietary metrics, computational resource demands, and the inability to exploit statistical relationships between Key Performance Indicators (KPIs, leading to delayed and incomplete problem detection.
Innovation Solution
Implement a two-stage process involving dimensionality reduction techniques, such as PCA, to transform high-dimensional network metrics into reduced dimensions, detect anomalies using supervised or unsupervised ML models, and map back to physical dimensions for actionable insights, using a GUI for visualization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional threshold-based KPI monitoring is used, then anomaly detection can be performed, but computational resources increase hugely and detection delays occur
Solution Approach 1:
The patent extracts and focuses only on the most critical KPIs that contribute to anomalies, rather than monitoring all KPIs equally. By identifying and extracting the subset of KPIs with highest anomaly contribution, the system reduces computational load while maintaining detection reliability
Solution Approach 2:
The patent segments the anomaly detection process into multiple stages: initial anomaly detection, contribution analysis, and root cause identification. This segmentation allows the system to process data efficiently at each stage rather than applying heavy computation uniformly across all data
2Reliability
If all raw performance metrics are monitored individually, then comprehensive coverage is achieved, but the complexity of analysis increases exponentially
Solution Approach 1:
The patent merges multiple related KPIs into composite indicators or aggregates them by functional groups. This combining approach maintains comprehensive monitoring coverage while reducing the number of individual metrics that require separate analysis, thereby simplifying the overall system complexity
Solution Approach 2:
The patent introduces a new dimension of analysis by examining KPI relationships and correlations rather than analyzing each metric in isolation. This dimensional shift from individual metric analysis to multi-metric relationship analysis reduces complexity while improving detection capability
3Measurement precision
If proprietary vendor-specific metrics are used, then vendor-specific performance is captured, but statistical relationships between KPIs cannot be exploited
Solution Approach 1:
The patent develops a universal analysis framework that can handle proprietary vendor-specific metrics while extracting common statistical patterns. The system is designed to work with metrics from different vendors by focusing on universal relationships and behaviors that transcend vendor-specific implementations
4Ease of manufacture
If traditional KPI threshold monitoring is applied, then simple implementation is achieved, but relationships between KPIs are not exploited leading to incomplete problem detection
Solution Approach 1:
The patent introduces an intermediary layer of analysis that sits between simple threshold monitoring and complex full-scale analysis. This intermediary layer captures KPI relationships through correlation analysis and contribution metrics, providing more complete problem detection while maintaining implementation feasibility
Data Source
AI summary
Systems and methods are provided for reducing a number of performance metrics generated by network functions to a number of reduced dimension metrics, which can be used to detect anomalous behavior and generate a warning signal of the detected anomalous behavior. The disclosed systems and methods transform raw performance metrics in a high dimensionality space to a reduced number of metrics in a lower dimensionality space through dimensionality reduction techniques. Anomalous behavior in network performance is detected in the high dimensionality space using the reduced dimension metrics. The systems and methods disclosed herein convert the reduced dimension metrics back to the high dimensionality space, such that the performance metrics from network functions can be utilized to understand and address potential problems in the network.