Graph Anomaly Detection With Graph Neural Networks: Current Status and Challenges

Kim, Hwan; Lee, Byung Suk; Shin, Won-Yong; Lim, Sungsu

doi:10.1109/access.2022.3211306

Cited by 39 publications

(10 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This precision in node detection surpasses traditional clustering methods, highlighting its criticality across diverse applications. For instance, in anomaly detection scenarios [51][52][53], the precise identification of unusual nodes is indispensable for tasks like fraud detection in financial transactions [54,55], intrusion detection in computer networks [56,57], and rare disease identification in biological networks [58,59]. Similarly, within network resource allocation frameworks [60,61]such as transportation or social networks, the ability to pinpoint nodes with specific characteristics is crucial for optimizing traffic flow, efficiently allocating resources, and upholding infrastructure integrity.…”

Section: -3-discussionmentioning

confidence: 99%

Efficient Graph Partitioning using LouvainSplit Algorithm

Javadi,

Aghighi,

Azadbakht

2024

Preprint

View full text Add to dashboard Cite

Graph partitioning is essential for uncovering cohesive communities within complex networks. This paper introduces LouvainSplit, an innovative algorithm designed to enhance graph partitioning efficiency and accuracy. LouvainSplit leverages advanced techniques in feature representation, community detection, and evaluation, providing a robust framework for addressing challenges inherent in graph partitioning tasks across diverse domains. At its core, LouvainSplit utilizes a feature pyramid representation approach to extract both basic and summary features from input graphs at multiple granularity levels. This methodology ensures a subtle evaluation of graph data by capturing fundamental graph information alongside intricate structural patterns, thus offering a comprehensive representation of underlying community structures. A key innovation of our approach is the integration of the Louvain algorithm, renowned for its efficacy in community detection. By leveraging pairwise cosine similarities computed from node feature vectors, the Louvain algorithm optimizes modularity iteratively, effectively partitioning graphs into cohesive communities. This iterative process facilitates the identification of significant network structures within complex networks, providing valuable insights into their organizational dynamics. To comprehensively evaluate LouvainSplit's performance, we integrated diverse graph partitioning algorithms, including PyMetis, Genetic Algorithm (GA), DBSCAN, KMeans, OPTICS, and spectral clustering to evaluate the effectiveness of LouvainSplit and using popular evaluation metrics across diverse benchmarks spanning various domains and graphs. The results demonstrate LouvainSplit's superiority in terms of recall score, modularity, and scalability compared to existing methods. Moreover, LouvainSplit maintains competitive average runtime values, ensuring efficient processing even with large-scale datasets.

show abstract

Section: -3-discussionmentioning

confidence: 99%

Efficient Graph Partitioning using LouvainSplit Algorithm

Javadi,

Aghighi,

Azadbakht

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…By learning the relationships between sensors, we could detect anomalies from sensors data [3][4][5] . However, traffic anomalies usually exhibit complex forms due to two aspects: high dimensionality, sparsity, abnormal scarcity (i.e., the need to correlate time and space, including speed or flow), and difficulty in capturing the hidden relationship between nodes (i.e., spatial modeling in the face of different data sources with varying degrees of anomalies in density or distribution and scale) 6,7 . Therefore, it is important to explore ways to capture complex inter-sensor relationships and detect anomalies from node relationships.…”

Section: Graph Autoencoder With Mirror Temporal Convolutional Network...mentioning

confidence: 99%

Graph autoencoder with mirror temporal convolutional networks for traffic anomaly detection

Ren,

Li,

Peng

et al. 2024

Sci Rep

View full text Add to dashboard Cite

Traffic time series anomaly detection has been intensively studied for years because of its potential applications in intelligent transportation. However, classical traffic anomaly detection methods often overlook the evolving dynamic associations between road network nodes, which leads to challenges in capturing the long-term temporal correlations, spatial characteristics, and abnormal node behaviors in datasets with high periodicity and trends, such as morning peak travel periods. In this paper, we propose a mirror temporal graph autoencoder (MTGAE) framework to explore anomalies and capture unseen nodes and the spatiotemporal correlation between nodes in the traffic network. Specifically, we propose the mirror temporal convolutional module to enhance feature extraction capabilities and capture hidden node-to-node features in the traffic network. Morever, we propose the graph convolutional gate recurrent unit cell (GCGRU CELL) module. This module uses Gaussian kernel functions to map data into a high-dimensional space, and enables the identification of anomalous information and potential anomalies within the complex interdependencies of the traffic network, based on prior knowledge and input data. We compared our work with several other advanced deep-learning anomaly detection models. Experimental results on the NYC dataset illustrate that our model works best compared to other models for traffic anomaly detection.

show abstract

“…With the emergence of Graph Neural Networks (GNNs), research on anomaly detection [21] has moved to spatio-temporal GNN (STGNN)-based approaches. GNNs offer a versatile framework for capturing and analyzing graphs' structural and attribute-based characteristics.…”

Section: Related Workmentioning

confidence: 99%

Statistically Inspired Discrepancy Detection for Anomalous Spatio-Temporal Graphs

Tiwari,

Chatterjee,

Padmakumar

et al. 2024

Preprint

View full text Add to dashboard Cite

Anomaly detection in dynamic graphs is a critical topic with applications in many fields, such as fraud detection and network security. This paper tackles the difficulties in locating abnormalities in time-varying graphs by presenting a novel divide-and-conquer method. We combine Graph Convolutional Networks (GCN) and Recurrent Neural Networks (RNN) to predict future node values on temporal graphs, followed by a macro and micro-level analysis. At the macro level, we present a novel algorithm to extract correlation-based subgraphs. The values obtained for each subgraph allow us to concentrate on subgraphs that show significant anomalies, effectively minimizing the challenges in node-level anomaly detection. This is followed by a micro-level analysis of the node contributions and temporal properties within the identified anomalous subgraphs. By combining macro and micro-level analysis with a machine learning-based approach, our method provides an efficient approach for zoning in on anomalies, significantly speeding up computation.

show abstract

Graph Anomaly Detection With Graph Neural Networks: Current Status and Challenges

Cited by 39 publications

References 52 publications

Efficient Graph Partitioning using LouvainSplit Algorithm

Efficient Graph Partitioning using LouvainSplit Algorithm

Graph autoencoder with mirror temporal convolutional networks for traffic anomaly detection

Statistically Inspired Discrepancy Detection for Anomalous Spatio-Temporal Graphs

Contact Info

Product

Resources

About