Network Service Anomaly Detection via Cross-Correlation Matrices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current anomaly detection methods in telecommunications networks are inefficient in identifying and troubleshooting issues across multiple domains due to their reliance on aggregated Key Performance Indicators (KPIs) and lack of consideration for the structural relationships between network functions, leading to delayed problem identification and manual troubleshooting processes.
Innovation Solution
A method that computes network service signatures as cross-correlation matrices from service-level and resource-level metrics across different domains, using auto-encoders to align time series data and a multi-modal machine learning model to detect anomalies based on reconstruction errors, enabling automated identification and classification of issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If anomaly detection methods rely on aggregated Key Performance Indicators (KPIs), then the monitoring process is simplified, but the precision of anomaly detection deteriorates due to loss of detailed structural relationships between network functions
Solution Approach 1:
The patent segments the network service into multiple network functions and further divides monitoring into service-level metrics and resource-level metrics. This segmentation allows detailed analysis of individual function relationships while maintaining overall system visibility, resolving the contradiction between simplified monitoring and precise detection.
Solution Approach 2:
The patent introduces a new dimensional framework by creating cross-correlation matrices that capture relationships across different metric dimensions (service-level and resource-level) and time dimensions. This multi-dimensional approach enables precise anomaly detection without requiring complex manual monitoring of each dimension separately.
2Ease of manufacture
If traditional anomaly detection methods are used, then implementation is easier, but the speed of problem identification deteriorates due to manual troubleshooting processes
Solution Approach 1:
The patent implements self-service anomaly detection through automated cross-correlation analysis and machine learning models that automatically identify anomalies and their root causes. The system serves itself by autonomously monitoring network functions, detecting deviations from normal cross-correlation patterns, and generating troubleshooting insights without manual intervention, thus reducing problem identification time while maintaining ease of deployment through automated processes.
3Measurement precision
If cross-correlations and long-term dependencies are captured in network services, then anomaly detection accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-computing cross-correlation matrices from historical network data and using machine learning models to learn normal correlation patterns before anomalies occur. This preparatory computation stores the complexity in advance, allowing rapid real-time anomaly detection without repeating heavy computational operations during actual monitoring, thus improving accuracy while managing computational complexity.
Data Source
AI summary
A method for anomaly detection and troubleshooting in a network includes parsing a network service descriptor (NSD) describing a network service (NS) to be deployed in the network. Monitoring data including time series of service-level metrics and resource-level metrics of network functions (NFs) of the NS are received from different domains of the network. Representations of the time series from the different domains are learned with a common dimensionality. An NS signature of the NS is computed as a cross-correlation matrix comprising cross-correlations between the service-level metrics and the resource-level metrics of the NFs. Embeddings of the NS signature are learned using a model and determining a reconstruction error of the model. It is determined whether the NS is anomalous based on the reconstruction error of the model. The NS is identified as a target for the troubleshooting in a case that the NS was determined to be anomalous.