DeepEC: Adversarial attacks against graph structure prediction models

Xian, Xingping; Wu, Tao; Qiao, Shaojie; Wang, Wei; Wang, Chao; Liu, Yanbing; Xu, Guangxia

doi:10.1016/j.neucom.2020.07.126

Cited by 20 publications

(3 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The successful portability of attacks against various link prediction heuristics from the existing literature is highlighted, demonstrating the effectiveness and broad applicability of the proposed competitive methods. Existing competing attacks (mentioned in [3,9,4,12]) often operate in a "white box" configuration, assuming full access to the machine learning model parameters. However, the passage notes the unrealistic nature of this setting when the model parameters are unknown to the attacker, emphasizing the importance of considering realistic scenarios, especially in the "black box" configuration.…”

Section: Generative Attacks Employ Reproductive Methods To Create Adv...mentioning

confidence: 99%

See 1 more Smart Citation

Generate Adversarial Attack on Graph Neural Network using K-Means Clustering and Class Activation Mapping

Ingle,

Pawale

2023

IJACSA

View full text Add to dashboard Cite

Graph Neural Networks (GNNs) have emerged as powerful tools for analyzing complex structured data, including social networks, biological networks, and recommendation systems. However, their susceptibility to adversarial attacks poses a significant challenge, especially in critical tasks such as node classification and link prediction. Adversarial attacks on GNNs can introduce harmful input graphs, leading to biased model predictions and compromising the integrity of the network. We propose a novel adversarial attack method that leverages the combination of K-Means clustering and Class Activation Mapping (CAM) to conduct subtle yet effective attacks against GNNs. The clustering algorithm identifies critical nodes within the graph, whose perturbations are likely to have a substantial impact on model performance. Additionally, CAM highlights regions of the graph that significantly influence GNN predictions, enabling more targeted and efficient attacks. We assess the efficacy of state-of-the-art GNN defenses against our proposed attack, underscoring the pressing need for robust defense mechanisms. Our study focuses on countering attacks on GNN networks by utilizing K-Means clustering and CAM to enhance the effectiveness and efficiency of the adversarial strategy. Through our observations, we emphasize the necessity for stronger security measures to safeguard GNN-based applications, particularly in sensitive environments. Furthermore, our research highlights the importance of developing robust GNNs that can withstand adversarial attacks, ensuring the reliability and trustworthiness of these models in critical applications. Strengthening the robustness of GNNs against adversarial manipulation is crucial for maintaining the security and integrity of systems that heavily rely on these advanced analytical tools. Our findings underscore the ongoing efforts required to fortify GNN-based applications, urging the research community and practitioners to collaborate in developing and implementing more robust security measures for these powerful neural network models. .

show abstract

Section: Generative Attacks Employ Reproductive Methods To Create Adv...mentioning

confidence: 99%

“…Carlini and Wagner [10,22] propose a method that comes close to the goal by utilizing simpler linear functions. [3] assumes linearity near the input, contributing to the minimal interference approach.…”

Section: Related Workmentioning

confidence: 99%

Generate Adversarial Attack on Graph Neural Network using K-Means Clustering and Class Activation Mapping

Ingle,

Pawale

2023

IJACSA

View full text Add to dashboard Cite

show abstract

“…It makes the predictive algorithms worse than random guesses. Xian et al [62] proposed a deep architecture-based framework to analyze the vulnerability of link prediction. The framework perturbed the network structure to adversarial attack the link prediction methods.…”

Section: Network Structural Perturbationmentioning

confidence: 99%

Network structural perturbation against interlayer link prediction

Tang¹,

Jiang²,

Chen³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Interlayer link prediction aims at matching the same entities across different layers of the multiplex network. Existing studies attempt to predict more accurately, efficiently, or generically from the aspects of network structure, attribute characteristics, and their combination. Few of them analyze the effects of intralayer links. Namely, few works study the backbone structures which can effectively preserve the predictive accuracy while dealing with a smaller number of intralayer links. It can be used to investigate what types of intralayer links are most important for correct prediction. Are there any intralayer links whose presence leads to worse predictive performance than their absence, and how to attack the prediction algorithms at the minimum cost? To this end, two kinds of network structural perturbation methods are proposed. For the scenario where the structural information of the whole network is completely known, we offer a global perturbation strategy that gives different perturbation weights to different types of intralayer links and then selects a predetermined proportion of intralayer links to remove according to the weights. In contrast, if these information cannot be obtained at one time, we design a biased random walk procedure, local perturbation strategy, to execute perturbation. Four kinds of interlayer link prediction algorithms are carried out on different real-world and artificial perturbed multiplex networks. We find out that the intralayer links connected with small degree nodes have the most significant impact on the prediction accuracy. The intralayer links connected with large degree nodes may have side

show abstract

Rethinking the Robustness of Graph Neural Networks

Cui

Zhao

et al. 2023

Communications in Computer and Information Science

View full text Add to dashboard Cite

DeepEC: Adversarial attacks against graph structure prediction models

Cited by 20 publications

References 51 publications

Generate Adversarial Attack on Graph Neural Network using K-Means Clustering and Class Activation Mapping

Generate Adversarial Attack on Graph Neural Network using K-Means Clustering and Class Activation Mapping

Network structural perturbation against interlayer link prediction

Rethinking the Robustness of Graph Neural Networks

Contact Info

Product

Resources

About