Skip to main navigation Skip to search Skip to main content

UnboundAttack: Generating Unbounded Adversarial Attacks to Graph Neural Networks

  • KTH Royal Institute of Technology
  • Laboratoire d'Informatique (LIX)

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Graph Neural Networks (GNNs) have demonstrated state-of-the-art performance in various graph representation learning tasks. Recently, studies revealed their vulnerability to adversarial attacks. While the available attack strategies are based on applying perturbations on existing graphs within a specific budget, proposed defense mechanisms successfully guard against this type of attack. This paper proposes a new perspective founded on unrestricted adversarial examples. We propose to produce adversarial attacks by generating completely new data points instead of perturbing existing ones. We introduce a framework, so-called UnboundAttack, leveraging the advancements in graph generation to produce graphs preserving the semantics of the available training data while misleading the targeted classifier. Importantly, our method does not assume any knowledge about the underlying architecture. Finally, we validate the effectiveness of our proposed method in a realistic setting related to molecular graphs.

Original languageEnglish
Title of host publicationComplex Networks and Their Applications XII - Proceedings of The 12th International Conference on Complex Networks and their Applications
Subtitle of host publicationCOMPLEX NETWORKS 2023 Volume 1
EditorsHocine Cherifi, Luis M. Rocha, Chantal Cherifi, Murat Donduran
PublisherSpringer Science and Business Media Deutschland GmbH
Pages100-111
Number of pages12
ISBN (Print)9783031534676
DOIs
Publication statusPublished - 1 Jan 2024
Event12th International Conference on Complex Networks and their Applications, COMPLEX NETWORKS 2023 - Menton, France
Duration: 28 Nov 202330 Nov 2023

Publication series

NameStudies in Computational Intelligence
Volume1141 SCI
ISSN (Print)1860-949X
ISSN (Electronic)1860-9503

Conference

Conference12th International Conference on Complex Networks and their Applications, COMPLEX NETWORKS 2023
Country/TerritoryFrance
CityMenton
Period28/11/2330/11/23

Keywords

  • Adversarial Attacks
  • Graph Neural Networks

Fingerprint

Dive into the research topics of 'UnboundAttack: Generating Unbounded Adversarial Attacks to Graph Neural Networks'. Together they form a unique fingerprint.

Cite this