Practical Attacks Against Graph-based Clustering

By Yizheng Chen, Yacin Nadji, Athanasios Kountouras, Fabian Monrose, Roberto Perdisci, Manos Antonakakis, Nikolaos Vasiloglou. 2017

90 citations recorded by the source corpus.

Graph modeling allows numerous security problems to be tackled in a general way, however, little work has been done to understand their ability to withstand adversarial attacks. We design and evaluate two novel graph attacks against a state-of-the-art network-level, graph-based detection system. Our work highlights areas in adversarial machine learning that have not yet been addressed, specifically: graph-based clustering techniques, and a global feature space where realistic attackers without perfect knowledge must be accounted for (by the defenders) in order to be practical. Even though less informed attackers can evade graph clustering with low cost, we show that some practical defenses are possible.

Subfields: Artificial Intelligence, Computer Networks and Communications, Statistical and Nonlinear Physics

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. Neil Zhenqiang Gong
2. Stephan Günnemann
3. Binghui Wang
4. Daniel Zügner
5. Lichao Sun
6. Philip S. Yu
7. Aleksandar Bojchevski
8. Bo Li

In Computer Networks and Communications

1. Binghui Wang
2. Neil Zhenqiang Gong
3. Daniel Zügner
4. Stephan Günnemann
5. Aleksandar Bojchevski
6. Victor C. M. Leung
7. Bo Li
8. Philip S. Yu

In Statistical and Nonlinear Physics

1. Daniel Zügner
2. Stephan Günnemann
3. Binghui Wang
4. Neil Zhenqiang Gong
5. Aleksandar Bojchevski
6. Amir Akbarnejad
7. Philip S. Yu
8. Michael K. Reiter

Citations in the corpus

• Attacking Graph-based Classification via Manipulating the Graph Structure (2019, background): “However, existing studies [7,8,12,18,19,21,27,30, 31,37,38,40,43,46,49,51–53,56,60,61,67,68,70] on adversarial machine learning mainly focused on machine learning for non-graph data, with only a few studies [15,16,59,74] as exceptions.”
• Adversarial Attacks on Neural Networks for Graph Data (2018, methodology): “For graph clustering, the work [9] has measured the changes in the result when injecting noise to a bi-partite graph that represent DNS queries.”
• Adversarial Attacks on Graph Neural Networks via Meta Learning (2019, background): “Chen et al. (2017) have measured the changes in the resulting graph clustering when injecting noise to a bi-partite graph that represent DNS queries.”
• A Survey on Security Threats and Defensive Techniques of Machine Learning: A Data Driven View (2018, background): “However, recently research [42] has shown there are two FIGURE 2.”
• Adversarial Attack and Defense on Graph Data: A Survey (2018, methodology): “ADR is a metric evaluating the defense performance according to the ASR defined above.”
• Adversarial Generative Nets: Neural Network Attacks on State-of-the-Art Face Recognition (2018, background): “Some work did not focus on physical-domain attacks per se, but rather digital-domain attacks that satisfy certain constraints [9], [22], [57], [64].”
• Adversarial Attacks on Node Embeddings via Graph Poisoning (2018, background): “In the context of graph clustering, Chen et al. (2017b) measure the performance changes when injecting noise to a bi-partite graph of DNS queries, but do not focus on automatically generating attacks.”