Interpretable and Effective Reinforcement Learning for Attacking against Graph-based Rumor Detection

TL;DR

This paper introduces an interpretable reinforcement learning-based attack method against graph neural network rumor detectors, utilizing domain-specific features and variance reduction techniques to improve attack effectiveness and interpretability.

Contribution

The paper presents a novel reinforcement learning framework with domain-specific features and variance reduction for interpretable, effective attacks on graph-based rumor detection models.

Findings

The proposed attack outperforms rule-based and end-to-end methods.

The credit assignment and variance reduction strategies improve training efficiency.

The attack policies are interpretable and effective in case studies.

Abstract

Social networks are frequently polluted by rumors, which can be detected by advanced models such as graph neural networks. However, the models are vulnerable to attacks and understanding the vulnerabilities is critical to rumor detection in practice. To discover subtle vulnerabilities, we design a powerful attacking algorithm to camouflage rumors in social networks based on reinforcement learning that can interact with and attack any black-box detectors. The environment has exponentially large state spaces, high-order graph dependencies, and delayed noisy rewards, making the state-of-the-art end-to-end approaches difficult to learn features as large learning costs and expressive limitation of graph deep models. Instead, we design domain-specific features to avoid learning features and produce interpretable attack policies. To further speed up policy optimization, we devise: (i) a credit assignment method that decomposes delayed rewards to atomic attacking actions proportional to the their camouflage effects on target rumors; (ii) a time-dependent control variate to reduce reward variance due to large graphs and many attacking steps, supported by the reward variance analysis and a Bayesian analysis of the prediction distribution. On three real world datasets of rumor detection tasks, we demonstrate: (i) the effectiveness of the learned attacking policy compared to rule-based attacks and current end-to-end approaches; (ii) the usefulness of the proposed credit assignment strategy and variance reduction components; (iii) the interpretability of the policy when generating strong attacks via the case study.