Graph Neural Networks (GNNs) have been shown to be susceptible to backdoor poisoning attacks, which pose serious threats to real-world applications. Meanwhile, graph reduction techniques have recently emerged as effective methods for accelerating GNN training on large-scale graphs. However, the development of graph reduction techniques for large graphs does not yet address how these methods might interact with the potential risks of data poisoning attacks against GNNs, particularly in relation to existing backdoor attacks. This paper conducts a thorough examination of the robustness of graph reduction methods in scalable GNN training in the presence of state-of-the-art backdoor attacks. We performed a comprehensive robustness analysis across six coarsening methods and six sparsification methods for graph reduction, under three GNN backdoor attacks against three GNN architectures. Our findings indicate that the effectiveness of graph reduction methods in mitigating attack success rates varies significantly, with some methods even exacerbating the attacks. Through detailed analyses of triggers and poisoned nodes, we interpret our findings and enhance our understanding of how graph reduction influences robustness against backdoor attacks. These results highlight the critical need for incorporating robustness considerations in graph reduction for scalable GNN training.
Vertical Federated Learning (VFL) is a federated learning paradigm where multiple participants, who share the same set of samples but hold different features, jointly train machine learning models. Although VFL enables collaborative machine learning without sharing raw data, it is still susceptible to various privacy threats. In this paper, we conduct the first comprehensive survey of the state-of-the-art in privacy attacks and defenses in VFL. We provide taxonomies for both attacks and defenses, based on their characterizations, and discuss open challenges and future research directions. Specifically, our discussion is structured around the model's life cycle, by delving into the privacy threats encountered during different stages of machine learning and their corresponding countermeasures. This survey not only serves as a resource for the research community but also offers clear guidance and actionable insights for practitioners to safeguard data privacy throughout the model's life cycle.
Iframes are widely used in modern web development, but their separation from the parent page can lead to iframe-specific rendering bugs, which occur only when content is embedded within an iframe. These bugs may degrade user experience and, in some cases, lead to serious security vulnerabilities. However, existing rendering bug detection techniques are largely ineffective since their oracles compare rendering outputs only within the same context and fail to capture discrepancies across different rendering contexts. This paper presents ICheck, an automated testing technique for iframe rendering in web browsers. Specifically, given an HTML document, ICheck introduces a cross-context oracle that detects inconsistencies by comparing the same HTML content in both top-level and iframe contexts. To further assist debugging, ICheck performs layout-aware input reduction by redefining the effective search space of bug-inducing elements. As a result, ICheck helps developers more effectively and efficiently identify and diagnose iframe-specific rendering bugs. Evaluated on the most popular browsers, Chrome and Firefox, ICheck detected 21 previously unknown bugs, all of which were confirmed. On average, ICheck uncovered 3.8× more bugs than state-of-the-art rendering bug detection work. In addition, ICheck effectively reduced HTML inputs by 93% in DOM nodes while achieving up to 10.1× faster reduction time compared with state-of-the-art delta debugging approaches.