ABSTRACT Background The rapid evolution of Internet of Things (IoT) technologies has driven innovations across domains such as robotics, autonomous systems, and environmental control. However, effectively learning graph‐based representations within these dynamic and heterogeneous systems remains a significant challenge, especially when scalability and adaptability are required. Aims This study aims to develop and evaluate a novel meta‐reinforcement learning (meta‐RL) framework that combines Deep Q‐Networks (DQNs) with Graph Convolutional Networks (GCNs) to learn adaptive and efficient representations of graph clusters. The primary objective is to enhance cluster‐based representation learning by integrating reinforcement learning with graph aggregation policies. Methods We propose a cluster policy‐GNN model that formulates optimal graph aggregation as a Markov Decision Process (MDP). The framework incorporates a cluster meta‐policy to guide node‐specific aggregation strategies and utilizes a combination of DQN and GCN for adaptive graph representation. Training involves clustering nodes based on policy‐determined hops and batching to ensure efficient GNN training. A custom reward function drives the reinforcement learning process to prioritize computational focus on the most informative subgraphs. Results Our experimental results, benchmarked on real‐world graph datasets, demonstrate that the proposed framework significantly outperforms existing state‐of‐the‐art methods, including static GNNs, alternating graph‐regularized networks, and causal‐aware neural architecture search models. The learned cluster policies effectively enhance representation learning by dynamically adjusting to the structural heterogeneity of input graphs. Improvements were observed across various domains and scales, validating the flexibility and generalizability of the method. Conclusion The proposed meta‐RL framework with integrated DQN and GCN modules offers a powerful and scalable approach for graph cluster representation learning. By introducing adaptive, node‐specific aggregation strategies guided by reinforcement learning, the method effectively captures complex graph structures and surpasses current techniques. Future work may explore real‐time adaptation and deployment in more dynamic IoT‐based applications.

Privacy inference poses a significant threat to users of online social networks (OSNs). To deal with this issue, a number of privacy-enhancing technologies have been proposed with the goal of achieving a balance between the protection of privacy and the utility of data. Previous studies, however, failed to take into consideration the impact of the interdependency of privacy (IoP), which dictates that privacy decisions made by some users may affect the privacy of some other users. The implication of IoP is that too much privacy may be disclosed when multiple individuals share data with the same data accessor because privacy conflicts resulting from independent privacy decisions would make it possible for adversaries to infer the privacy of the target user. Ideally, cooperation that preserves privacy should allow OSN users to respect each other’s privacy specifications so as to resolve such privacy conflicts caused by independent privacy decisions of individuals. To facilitate the design, we propose a privacy-preserving cooperation framework based on the evolutionary game theory to facilitate such cooperation. Based on the framework, the dynamics of user strategies regarding whether to participate in the cooperation are analyzed and an evolutionary stable state is derived to serve as the basis for incentivizing users to participate in cooperative privacy protection. Experiments based on real OSN data show that the proposed cooperation framework is effective in modeling the behaviors of users and that the proposed incentive allocation method can incentivize users to participate in the cooperation. The proposed cooperation framework can not only helps lower the threat to user privacy resulting from privacy inference by data accessors but also allows OSN service providers to design effective privacy protection policies.

Privacy protection is a hot topic in network security, many scholars are committed to evaluating privacy information disclosure by quantifying privacy, thereby protecting privacy and preventing telecommunications fraud. However, in the process of quantitative privacy, few people consider the reasoning relationship between privacy information, which leads to the underestimation of privacy disclosure and privacy disclosure caused by malicious reasoning. This paper completes an experiment on privacy information disclosure in the real world based on WordNet ontology .According to a privacy measurement algorithm, this experiment calculates the privacy disclosure of public figures in different fields, and conducts horizontal and vertical analysis to obtain different privacy disclosure characteristics. The experiment not only shows the situation of privacy disclosure, but also gives suggestions and method to reduce privacy disclosure.

Collaborative filtering based on single domains has become widely used in today's recommendation system. Nevertheless, it has two problems that need to be solved, i.e., the cold start problem and the data sparseness problem. As the result, cross-domain recommendation technology has emerged, which aims at integrating user preference characteristics from different domains. This paper proposes a collaborative filtering recommendation method based on multi-domain semantic fusion (CF-MDS). CF-MDS achieves cross-domain item similarity calculation through semantic analysis and ontology and integrates data from different domains iteratively based on domain relevance to rate users on target domain items and to produce a cross-domain user-item rating matrix. Collaborative filtering technology is then combined with multi-domain fusion recommendation algorithm. Experimental results show that the proposed method can deal effectively with the cold start problem and data sparsity problem that exist in traditional recommendation systems as well as can improve the diversity of recommendation. Compared to other cross-domain recommendation methods, the proposed method can better meet personal needs of users and also improve the accuracy of recommendation.