Deep Anomaly Detection on Attributed Networks

By Kaize Ding, Jundong Li, Rohit Bhanushali, Huan Liu. SDM, 2019

630 citations recorded by the source corpus.

Attributed networks are ubiquitous and form a critical component of modern information infrastructure, where additional node attributes complement the raw network structure in knowledge discovery. Recently, detecting anomalous nodes on attributed networks has attracted an increasing amount of research attention, with broad applications in various high-impact domains, such as cybersecurity, fi-nance, and healthcare. Most of the existing attempts, however, tackle the problem with shallow learning mechanisms by ego-network or community analysis, or through subspace selection. Undoubtedly, these models cannot fully address the computational challenges on attributed networks. For example, they often suffer from the network sparsity and data nonlinearity issues, and fail to capture the complex interactions between different information modalities, thus negatively impact the performance of anomaly detection. To tackle the aforementioned problems, in this paper, we study the anomaly detection problem on attributed networks by developing a novel deep model. In particular, our proposed deep model: (1) explicitly models the topological structure and nodal attributes seamlessly for node embedding learning with the prevalent graph convolutional network (GCN); and (2) is customized to address the anomaly detection problem by virtue of deep autoencoder that leverages the learned embeddings to reconstruct the original data. The synergy between GCN and autoencoder enables us to spot anomalies by measuring the reconstruction errors of nodes from both the structure and the attribute perspectives. Extensive experiments on real-world attributed network datasets demonstrate the efficacy of our proposed algorithm.

Subfields: Artificial Intelligence, Molecular Biology, Statistical and Nonlinear Physics

View Paper: https://www.semanticscholar.org/paper/bcfdbb6b8911272139170ef4b24e31d9145093e7

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Huan Liu
2. Chuan Zhou
3. Chunhua Shen
4. Anton van den Hengel
5. Shirui Pan
6. Philip S. Yu
7. Jundong Li
8. Yixin Liu

In Molecular Biology

1. Chuan Zhou
2. Philip S. Yu
3. Huan Liu
4. Lu Cheng
5. P. Richard Hahn
6. Jia Wu
7. Quan Z. Sheng
8. Chunhua Shen

In Statistical and Nonlinear Physics

1. Chuan Zhou
2. Huan Liu
3. Shirui Pan
4. Yixin Liu
5. Jundong Li
6. Kaize Ding
7. George Karypis
8. Chen Gong

Citations in the corpus

• Anomaly Detection on Attributed Networks via Contrastive Self-Supervised Learning (2021, methodology): “DOMINANT [13] is one of the representative methods.”
• Graph Self-Supervised Learning: A Survey (2021, methodology): “To be concrete, DOMINANT [120], SpecAE [121] and AEGIS [122] employ hybrid SSL frameworks that combine structure and feature generation to capture the patterns of anomalies.”
• Deep Learning for Anomaly Detection (2020, methodology): “Different types of neural network layers and architectures can be used under the AE framework, allowing us to detect anomalies in high-dimensional data, as well as non-independent data such as attributed graph data [37] and multivariate sequence data [91, 97] (CH2).”
• Rethinking Graph Neural Networks for Anomaly Detection (2022, methodology): “To control α , we follow the attribute perturbation schema in (Ding et al., 2019).”
• Next-item Recommendation with Sequential Hypergraphs (2020, background): “There is an increasing attention on exploiting graph structures for various recommendation scenarios with the recent advance in neural graph embedding algorithms [8, 9, 13, 19, 34].”
• Graph Few-shot Class-incremental Learning (2021, background): “Graph Neural Network (GNN) [1, 3, 9, 10, 16, 41, 45] is a family of deep neural models tailored for graph-structured data, which has been widely used in various applications, such as recommendation [42], anomaly detection [4], and text classification [5].”
• A Comprehensive Survey on Graph Anomaly Detection With Deep Learning (2021, methodology): “Lastly, ALARM adopts a similar scoring function as [92], and nodes with top-k highest scores are anomalous.”
• A Survey of Learning Causality with Data (2018, background): “Potential problems include handling (1) anomalies [1, 37], (2) treatment entanglement [151], (3) complex treatments [95] (e.g., image and text as treatment), (4) temporal observations [102, 132].”