Reinforcement Learning For Data Poisoning on Graph Neural Networks
This addresses a security issue for users of graph classification systems, but it is incremental as it extends existing adversarial methods to a less-explored setting.
The paper tackles the problem of data poisoning attacks on graph neural networks for graph classification, proposing a reinforcement learning approach and achieving results that demonstrate the vulnerability of such models.
Adversarial Machine Learning has emerged as a substantial subfield of Computer Science due to a lack of robustness in the models we train along with crowdsourcing practices that enable attackers to tamper with data. In the last two years, interest has surged in adversarial attacks on graphs yet the Graph Classification setting remains nearly untouched. Since a Graph Classification dataset consists of discrete graphs with class labels, related work has forgone direct gradient optimization in favor of an indirect Reinforcement Learning approach. We will study the novel problem of Data Poisoning (training time) attack on Neural Networks for Graph Classification using Reinforcement Learning Agents.