Robust Heterogeneous Graph Neural Networks against Adversarial Attacks

Zhang, Mengmei; Wang, Xiao; Zhu, Meiqi; Shi, Chuan; Zhang, Zhiqiang; Zhou, Jun

doi:10.1609/aaai.v36i4.20357

Cited by 27 publications

(8 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, some studies have been conducted on developing robust GNN models [27][28][29][30][31][32]. This section discusses the use of robustness during embedding graphs, including adversarial attacks and defenses.…”

Section: Related Workmentioning

confidence: 99%

Robust graph representation learning via out-of-distribution detection approach

Bastami,

Soltanizadeh,

Rahmanimanesh

et al. 2024

Preprint

View full text Add to dashboard Cite

Graph neural networks (GNNs) are powerful models capable of learning from graph-structured data and performing various tasks. GNNs are susceptible to poisoning attacks, in which sophisticated attackers inject malicious nodes or edges into the graph topology to degrade their performance. The existing defense mechanisms, such as adversarial training, are ineffective in improving the robustness of GNN models. Fake nodes can be utilized creatively to deceive traditional GNN neural functions. In this paper, we propose a robust GNN model empowered by a reliable aggregation function based on the OOD detection concept and a novel method. A key idea of RGRO is to train a model that maps the nodes to a latent space in which we can measure the distance between the nodes and their distribution. The Mahalanobis distance is proposed as a superior alternative to cosine distance in order to exploit the homophily rule better and to consider the contextual information of the nodes. The Mahalanobis distance, which considers the covariance of the data and is scale-invariant, allows for a more effective application of the homophily rule. Mahalanobis distance is explored in this study to enhance the robustness and accuracy of outlier detection in graph data, thus contributing to a better understanding of homophily. The RGRO model can improve accuracy by removing poisoned data without knowing any types of poisoning attacks or GNN algorithms. RGRO is evaluated with four typical defense strategies based on two types of poisoning attacks implemented on different realistic datasets. Based on the results, RGRO can detect poisoned data against attacks effectively and efficiently. In the best scenario, RGRO improves the accuracy of the GNN model by 0.86.

show abstract

Section: Related Workmentioning

confidence: 99%

Robust graph representation learning via out-of-distribution detection approach

Bastami,

Soltanizadeh,

Rahmanimanesh

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…In this approach, the representation of the central node is iteratively learned and enhanced by amalgamating and transforming the representations of its adjacent nodes. These networks play a significant role in network analysis and learning [15][16][17]. Among various GNN models, the Graph Convolutional Network (GCN), introduced by Kipf et al [18], is particularly noteworthy.…”

Section: B Network Analysis and Learning Based On Gnnmentioning

confidence: 99%

A Graph Attention Network-Based Link Prediction Method Using Link Value Estimation

Zhang,

Wu,

Zhu

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Link prediction in complex networks is a critical process aimed at uncovering hidden or potential connections among nodes. This technique is widely utilized in areas such as knowledge graphs. Current Graph Neural Networks (GNNs) often focus exclusively on determining whether nodes are connected or assessing the strength of these links by leveraging node attributes. They typically use network structure and attributes to develop node representations through neighborhood aggregation. However, these methods often overlook the intrinsic importance of the links themselves. This paper thoroughly examines the significance of link value based on network structure and introduces an innovative approach for estimating this value, and proposes a method that incorporates link value into both the formulation and training of a link prediction graph attention network. This integration not only boosts the accuracy of link predictions but also provides a theoretical basis for understanding the prediction results. We conducted extensive experiments in link prediction employing widely recognized benchmark datasets. The findings reveal that our proposed framework for link prediction exhibits commendable performance and generalization capabilities, and overall performance improved by an average of 1.2%, thereby establishing it as an effective baseline model.

show abstract

“…In particular, the threats are (1) these representations reveal sensitive attributes, no matter if they explicitly exist in the input text and (2) the representations can be partially recovered via generative models. In our recent paper (Zhan et al 2023), we propose a GRA to recover a graph's adjacency matrix from three types of representation outputs, i.e., representation outputs from graph convolutional networks, graph attention networks, and SNNs. We find that SNN outputs obtain the highest precision and AUC on five real-world networks.…”

Section: Research Planmentioning

confidence: 99%

Privacy-Preserving Representation Learning for Text-Attributed Networks with Simplicial Complexes

Zhan¹,

Sheng²

2023

AAAI

View full text Add to dashboard Cite

Although recent network representation learning (NRL) works in text-attributed networks demonstrated superior performance for various graph inference tasks, learning network representations could always raise privacy concerns when nodes represent people or human-related variables. Moreover, standard NRLs that leverage structural information from a graph proceed by first encoding pairwise relationships into learned representations and then analysing its properties. This approach is fundamentally misaligned with problems where the relationships involve multiple points, and topological structure must be encoded beyond pairwise interactions. Fortunately, the machinery of topological data analysis (TDA) and, in particular, simplicial neural networks (SNNs) offer a mathematically rigorous framework to evaluate not only higher-order interactions, but also global invariant features of the observed graph to systematically learn topological structures. It is critical to investigate if the representation outputs from SNNs are more vulnerable compared to regular representation outputs from graph neural networks (GNNs) via pairwise interactions. In my dissertation, I will first study learning the representations with text attributes for simplicial complexes (RT4SC) via SNNs. Then, I will conduct research on two potential attacks on the representation outputs from SNNs: (1) membership inference attack, which infers whether a certain node of a graph is inside the training data of the GNN model; and (2) graph reconstruction attacks, which infer the confidential edges of a text-attributed network. Finally, I will study a privacy-preserving deterministic differentially private alternating direction method of multiplier to learn secure representation outputs from SNNs that capture multi-scale relationships and facilitate the passage from local structure to global invariant features on text-attributed networks.

show abstract

Robust Heterogeneous Graph Neural Networks against Adversarial Attacks

Cited by 27 publications

References 18 publications

Robust graph representation learning via out-of-distribution detection approach

Robust graph representation learning via out-of-distribution detection approach

A Graph Attention Network-Based Link Prediction Method Using Link Value Estimation

Privacy-Preserving Representation Learning for Text-Attributed Networks with Simplicial Complexes

Contact Info

Product

Resources

About