The rapid spread of multimodal fake news poses a serious societal threat, as its evolving nature and reliance on timely factual details challenge existing detection methods. Dynamic Retrieval-Augmented Generation provides a promising solution by triggering keyword-based retrieval and incorporating external knowledge, thus enabling both efficient and accurate evidence selection. However, it still faces challenges in addressing issues such as redundant retrieval, coarse similarity, and irrelevant evidence when applied to deceptive content. In this paper, we propose ExDR—an Explanation-driven Dynamic Retrieval-Augmented Generation framework for Multimodal Fake News Detection. Our framework systematically leverages model-generated explanations in both the retrieval triggering and evidence retrieval modules. It assesses triggering confidence from three complementary dimensions, constructs entity-aware indices by fusing deceptive entities, and retrieves contrastive evidence based on deception-specific features to challenge the initial claim and enhance the final prediction. Experiments on two benchmark datasets, AMG and MR2, demonstrate that ExDR consistently outperforms previous methods in retrieval triggering accuracy, retrieval quality, and overall detection performance, highlighting its effectiveness and generalization capability.

Data protection laws and regulations require service providers to disclose data practices in privacy policies, specifying what personal information is processed and for what purposes. For compliance, these data practices must adhere to the data minimization principle, limiting the processing of personal information to what is directly relevant and necessary for the service purposes. However, data minimization is context-dependent, making violations difficult to define and quantify in privacy policies. Meanwhile, privacy policies are semantically complex and unstructured, hindering accurate extraction of fine-grained data practices and large-scale automated evaluation. To address these issues, we propose DataMini, a human--LLM collaborative evaluation framework for identifying violations of the data minimization principle in privacy policies. First, DataMini categorizes data minimization violations into two dimensions: inherent violations and contextual violations, establishing fine-grained evaluation criteria. Second, we construct a compliance baseline by mining high-frequency patterns from large-scale privacy policies and integrating expert knowledge to derive compliance mappings for human--LLM collaborative evaluation. Finally, the compliance baseline can automatically verify data practices that satisfy the data minimization principle, enabling the framework to focus exclusively on identifying suspected violations to improve efficiency and accuracy. Extensive evaluations demonstrate that DataMini exhibits superior data practice extraction accuracy of 83.46% and achieves an F1-score of 0.8180 for identifying data minimization violations in privacy policies, reducing manual evaluation effort by approximately 80%.

Access control policies play a critical role in securing sensitive data and protecting personal rights in environments such as cloud computing and IoT. These policies, typically created by sysadmins, specify which users are authorized to access specific resources under certain conditions. However, the manual creation and revision of these policies to align with security objectives is often error-prone and labor-intensive. In this paper, we present Custominer, a policy mining tool designed to assist sysadmins in proactively generating and customizing access control policies that meet predefined security requirements. Custominer enables sysadmins to define security goals as constraints, and then automatically mines policies that satisfy these constraints from access logs. The policy mining task is framed as a local search optimization problem, utilizing a MaxSAT solver to efficiently eliminate suboptimal policy candidates. Our experiments, conducted on four real-world datasets, show that Custominer outperforms existing state-of-the-art methods in terms of both accuracy and efficiency.

As cross-domain access constitutes a significant portion of network communication, multi-domain networks present both enhanced capabilities and increased cybersecurity risks. Traditional defense strategies often overlook the complexities of cross-domain collaboration, particularly the strategic interactions among domains that prioritize their own interests. In this paper, we introduce Macd, a multi-domain collaborative defense framework, which models the defense interactions as a multi-agent stochastic game. This enables Macd to consider long-term security performance across domains, mitigating multi-step attack threats. To promote effective collaboration, we propose a Shapley-value based reputation mechanism to ensure fair incentives for non-attacked domains that contributing Security Service Functions (SSFs). Additionally, we implement a MAPPO-based Macd-solver to dynamically compute optimal defense strategies. Simulations in a DDoS attack-defense scenario demonstrate that Macd significantly enhances cross-domain collaboration and improves the overall security of multi-domain networks.

In an increasingly interconnected world, multi-domain networks serve as vital infrastructure, enabling seamless communication and resource sharing across diverse sectors, but also leading to increasingly frequent cyberattacks. Defense-in-depth (DiD) is widely regarded as a necessary strategy for mitigating these threats through layered security measures. However, current DiD strategies often fall short due to their single-domain focus, reliance on centralized control, and inability to adapt to dynamic threats. This paper proposes a novel framework to extend DiD strategies for multi-domain networks. It progressively defines key elements of multi-domain networks, culminating in a detailed hierarchical framework that clarifies the roles and interactions of management domains. Furthermore, cross-domain intrusion response is modeled as a multi-agent stochastic game, accounting for self-interested behavior and interactions between domains. The Independent Q-Learning (IQL) algorithm is employed to solve this game, with experimental results demonstrating substantial improvements in security across multi-domain environments.