Graph Anomaly Detection via Statistical Similarity Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In large data scenarios, it is impractical to manually analyze structured data, such as graphs or trees, to detect anomalies, which can involve deviations in node labels and structural relationships, making automated detection necessary.
Innovation Solution
The method involves extracting Structured and Labeled (SL) features from graphs, constructing a consensus graph based on training data, and using a statistical model to compute similarity scores for examined graphs, identifying anomalies by comparing their features to the reference model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automated detection methods are used to analyze structured data in large data scenarios, then productivity and efficiency are improved, but measurement precision and detection accuracy may deteriorate due to the complexity of automated algorithms
Solution Approach 1:
The patent segments the anomaly detection process into distinct modules: graph feature extraction module that identifies structural patterns, statistical model module that establishes baseline distributions, and scoring module that quantifies anomaly likelihood. This segmentation allows each module to specialize in specific tasks, improving overall detection accuracy while maintaining high throughput for large datasets
Solution Approach 2:
The patent introduces a statistical model as an intermediary between the raw graph data and the anomaly detection decision. The statistical model learns normal patterns from training data and serves as a reference baseline, enabling the system to accurately identify deviations without requiring complex manual analysis rules, thus maintaining precision at scale
2Measurement precision
If manual analysis is used to detect anomalies in structured data, then measurement precision is improved, but productivity and time consumption deteriorate
Solution Approach 1:
The patent performs preliminary action by pre-training statistical models on normal graph data before actual anomaly detection. The system pre-extracts graph features and establishes baseline distributions during a training phase, so that during production, anomaly detection can be performed rapidly by comparing against pre-computed statistics, achieving both speed and accuracy
3Measurement precision
If complex statistical models are used to analyze graph structures, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent applies local quality by focusing statistical analysis on specific local graph features (such as node degrees, path lengths, subgraph patterns) rather than attempting to analyze the entire graph structure globally. This localized approach maintains detection precision for relevant anomalies while reducing computational complexity and resource requirements
Data Source
AI summary
A method, apparatus and product for automatic detection of anomalies in graphs. The method comprising obtaining training data, the training data comprising a plurality of graphs, each defined by nodes and edges connecting between the nodes, at least some of the nodes are labeled; determining a statistical model of a graph in accordance with the training data, the statistical model takes into account at least one structured and labeled feature of the graph, wherein the structured and labeled feature of the graph is defined based on a connection between a plurality of nodes and based on at least a portion of the labels of the plurality of nodes; obtaining an examined graph; and determining a score of the examined graph indicative of a similarity between the examined graph and the training data, wherein the score is based on a value of the structured and labeled feature in the examined graph.