Graph Neural Network for Dynamic Network Risk Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network risk prediction methods for dynamic networks are inefficient and fail to extract both structure and attribute features, leading to low accuracy and inability to detect anomaly changes, which hampers early deployment of network defense.
Innovation Solution
A dynamic network risk predicting method using a graph neural network that models time sequence feature graphs, incorporating long-short term memory networks to capture anomaly changes, and employs graph convolution neural networks for feature extraction and a robust random cutting forest algorithm for anomaly detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional risk predicting methods are used, then the prediction process is simple, but the accuracy of network anomaly risk prediction is low and efficiency is poor
Solution Approach 1:
The patent combines multiple neural network models (Graph Neural Network, LSTM, Random Forest) into a composite prediction system. Each component addresses specific limitations: GNN extracts structural features from network relationships, LSTM captures temporal dynamics, and Random Forest provides robust anomaly classification. This composite architecture resolves the contradiction by achieving high accuracy through feature fusion while managing complexity through modular design.
Solution Approach 2:
The prediction system is segmented into distinct functional modules: network data acquisition and graph construction, GNN-based structural feature extraction, LSTM-based temporal feature extraction, and Random Forest-based anomaly classification. This segmentation allows each component to specialize in specific tasks, improving overall accuracy while making the complex system manageable through clear separation of concerns.
2Measurement precision
If existing predicting methods are used, then the implementation is straightforward, but they cannot extract both structure features and attribute features simultaneously
Solution Approach 1:
The Graph Neural Network is designed with multi-functionality to simultaneously extract both structural features (from graph topology) and attribute features (from node/edge properties). The GNN architecture integrates structural information through message passing between connected nodes while incorporating attribute information through node feature transformations. This universal feature extraction capability resolves the contradiction by achieving comprehensive feature extraction in a unified model rather than requiring separate processing pipelines.
3Reliability
If conventional methods are used, then the system is easy to deploy, but they do not take into account anomaly changes in the network during the prediction process
Solution Approach 1:
The patent introduces Long Short-Term Memory (LSTM) networks to capture temporal dynamics and evolution of network anomalies. Unlike static conventional methods, the LSTM model processes sequential network data to identify how anomalies change over time, capturing patterns such as escalating attack behaviors or evolving threat scenarios. This dynamic modeling approach resolves the contradiction by enabling detection of anomaly changes while managing complexity through the specialized temporal processing capabilities of LSTM architecture.
Data Source
AI summary
Disclosed is a dynamic network risk predicting method based on a graph neural network, which comprises the following steps: collecting network data firstly and constructing a network data sequence graph; extracting the time sequence of the network data sequence graph and obtaining a time sequence feature graph; extracting network attribute features, network structure features and network change features of the dynamic network; obtaining the representation vector of the network sequence graph by learning the representation vector; constructing the anomaly detection model and carrying out the dynamic network risk prediction.