About this Research Topic
Graph-based machine learning has been widely applied to study security and privacy problems, including fraud detection (e.g., malware detection, fake user/review detection in social networks, malicious website detection, auction fraud detection), APT infection detection, attribute inference attacks, etc. However, graph-based machine learning methods are also demonstrated to be vulnerable when facing an adversary. Existing studies on ma-chine learning security and privacy mainly focused on machine learning for non-graph data, and security and privacy in graph-based machine learning is largely unexplored.
The goal of this Research Topic is to make graph-based machine learning secure and private. To this end, we encourage researchers to provide solutions to address security and privacy issues faced by graph-based machine learning methods (especially for collective classification methods and graph neural network methods). Important problems include:
• Security attacks (e.g., evasion attacks, poisoning attacks, backdoor attacks) to graph-based machine learning
• Empirical/certified defense against security attacks to graph-based machine learning
• Privacy attacks (e.g., membership inference, model inversion, attribute inference, proper-ty inference, etc.) to graph-based machine learning
• Empirical/certified defense against privacy attacks to graph-based machine learning
• Privacy-preserving graph-based machine learning
• Interpretable graph-based machine learning
• Theoretical understanding of the connection between accuracy, robustness, or/and privacy
Keywords: Graph-based machine learning, Privacy-preserving graph-based machine learning, Security, Privacy, Adversarial graph-based machine learning, Security attacks, Privacy attacks
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.