Abstract:While deep neural networks have achieved great success on the graph analysis, recent works have shown that they are also vulnerable to adversarial attacks where fraudulent users can fool the model with a limited number of queries. Compared with adversarial attacks on image classification, performing adversarial attack on graphs is challenging because of the discrete and non-differential nature of a graph. To address these issues, we proposed Cluster Attack, a novel adversarial attack by introducing a set of fake nodes to the original graph which can mislead the classification on certain victim nodes. Specifically, we query the victim model for each victim node to acquire their most adversarial feature, which is related to how the fake node’s feature will affect the victim nodes. We further cluster the victim nodes into several subgroups according to their most adversarial features such that we can reduce the searching space. Moreover, our attack is performed in a practical and unnoticeable manner: (1) We protect the predicted labels of nodes which we are not aimed for from being changed during attack. (2) We attack by introducing fake nodes into the original graph without changing existing links and features. (3) We attack with only partial information about the attacked graph, i.e., by leveraging the information of victim nodes along with their neighbors within k-hop instead of the whole graph. (4) We perform attack with a limited number of queries about the predicted scores of the model in a blackbox manner, i.e., without model architecture and parameters. Extensive experiments demonstrate the effectiveness of our method in terms of the success rate of attack. Department of Computer Science and Technology, Tsinghua University, China. Correspondence to: Zhengyi Wang <wangzy21@mails.tsinghua.edu.cn>. x x x

Attacking Community Detectors: Mislead Detectors via Manipulating the Graph Structure

Blindfolded Attackers Still Threatening: Strict Black-Box Adversarial Attacks on Graphs.

Ga-Based Q-Attack On Community Detection

Devil in Disguise: Breaching Graph Neural Networks Privacy through Infiltration

ASA: Adversary Situation Awareness Via Heterogeneous Graph Convolutional Networks.

Multiscale Evolutionary Perturbation Attack on Community Detection

Using Overlapping Methods to Counter Adversaries in Community Detection

From Bi-Level to One-Level: A Framework for Structural Attacks to Graph Anomaly Detection

A General Black-box Adversarial Attack on Graph-based Fake News Detectors

Unified Robust Network Embedding Framework for Community Detection via Extreme Adversarial Attacks

Towards Robust Community Detection Via Extreme Adversarial Attacks.

DMGNN: Detecting and Mitigating Backdoor Attacks in Graph Neural Networks

Attacking Fake News Detectors via Manipulating News Social Engagement

Certified Robustness of Community Detection against Adversarial Structural Perturbation via Randomized Smoothing

DeepInsight: Topology Changes Assisting Detection of Adversarial Samples on Graphs

Swarm intelligence for protecting sensitive identities in complex networks

BinarizedAttack: Structural Poisoning Attacks to Graph-based Anomaly Detection

A Qualitative Survey on Community Detection Attack Algorithms

Query-based Adversarial Attacks on Graph with Fake Nodes

Adversarial Detection on Graph Structured Data.

Botnet Detection using Social Graph Analysis