A Unified Framework for Data Poisoning Attack to Graph-based Semi-supervised Learning

TL;DR

This paper introduces a comprehensive framework for data poisoning attacks on graph-based semi-supervised learning, providing specialized algorithms for regression and classification, and demonstrating significant vulnerability in real datasets.

Contribution

The paper unifies various data poisoning attack scenarios into a single framework and develops efficient algorithms for both regression and classification cases in G-SSL.

Findings

Poisoning two labeled data points can cause MNIST classifier to perform at chance level.

The proposed algorithms effectively find optimal or near-optimal poisoning perturbations.

The framework demonstrates the vulnerability of G-SSL models to data poisoning attacks.

Abstract

In this paper, we proposed a general framework for data poisoning attacks to graph-based semi-supervised learning (G-SSL). In this framework, we first unify different tasks, goals, and constraints into a single formula for data poisoning attack in G-SSL, then we propose two specialized algorithms to efficiently solve two important cases --- poisoning regression tasks under $\ell_2$-norm constraint and classification tasks under $\ell_0$-norm constraint. In the former case, we transform it into a non-convex trust region problem and show that our gradient-based algorithm with delicate initialization and update scheme finds the (globally) optimal perturbation. For the latter case, although it is an NP-hard integer programming problem, we propose a probabilistic solver that works much better than the classical greedy method. Lastly, we test our framework on real datasets and evaluate the robustness of G-SSL algorithms. For instance, on the MNIST binary classification problem (50000 training data with 50 labeled), flipping two labeled data is enough to make the model perform like random guess (around 50\% error).