English

A General Framework for Defending Against Backdoor Attacks via Influence Graph

Machine Learning 2021-11-30 v1 Artificial Intelligence Computation and Language Computer Vision and Pattern Recognition

Abstract

In this work, we propose a new and general framework to defend against backdoor attacks, inspired by the fact that attack triggers usually follow a \textsc{specific} type of attacking pattern, and therefore, poisoned training examples have greater impacts on each other during training. We introduce the notion of the {\it influence graph}, which consists of nodes and edges respectively representative of individual training points and associated pair-wise influences. The influence between a pair of training points represents the impact of removing one training point on the prediction of another, approximated by the influence function \citep{koh2017understanding}. Malicious training points are extracted by finding the maximum average sub-graph subject to a particular size. Extensive experiments on computer vision and natural language processing tasks demonstrate the effectiveness and generality of the proposed framework.

Keywords

Cite

@article{arxiv.2111.14309,
  title  = {A General Framework for Defending Against Backdoor Attacks via Influence Graph},
  author = {Xiaofei Sun and Jiwei Li and Xiaoya Li and Ziyao Wang and Tianwei Zhang and Han Qiu and Fei Wu and Chun Fan},
  journal= {arXiv preprint arXiv:2111.14309},
  year   = {2021}
}
R2 v1 2026-06-24T07:55:09.535Z