JTJinchuan Tang

Jinchuan Tang

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Jinchuan Tang has a total of 30 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Entropy-Driven Action Randomization: A Deployment-Time Defense for Environment Privacy in Deep Reinforcement Learning

    by Xin Cheng, Jinchuan Tang, Shuping Dang

    Deep reinforcement learning (DRL) agents implicitly memorize the structure of their training environment, allowing an attacker to reconstruct it by querying the action outputs of a deployed model and causing environment-privacy leakage. To address this, this study proposes Entropy-Driven Action Randomization (EDAR), a deployment-time action randomization defense. Inspired by the exponential mechanism from differential privacy, EDAR replaces deterministic greedy action selection with probabilistic action sampling based on a normalized Q-value utility function, without altering the trained policy. A state-adaptive dynamic privacy budget is further designed, guided by the Shannon entropy of the action distribution. A simulated-annealing-based policy inversion attack is used to quantify privacy leakage, measured by the normalized recovery rate (NRR). Experiments on GridWorld environments ranging from 7 × 7 to 13 × 13 show that, at a privacy level comparable to a strong fixed budget, the proposed dynamic budget raises the reward retention from 6.2% to 78.9% while keeping a comparable NRR reduction. These results indicate that adapting the privacy budget to per-state policy determinism yields a markedly better privacy–utility trade-off than existing training-perturbation-based methods. We clarify that the differential privacy property invoked here is a per-query, mechanism-level guarantee of indistinguishability between neighboring observation states under a fixed trained model. The protection of the environment structure itself is established empirically through the policy inversion attack. Experiments are conducted on discrete GridWorld navigation tasks; the conclusions are scoped to small value-based navigation settings.


  • Open Access

    A Conditional Diffusion Graph Privacy Protection Model Based on Task-Aware Attention and Hierarchical Adversarial Constraints

    by Yifan Wang, Jinchuan Tang, Shuping Dang

    Graph representation learning models inherently capture structural and attribute correlations, which may lead to significant privacy risks under attribute inference attacks. Existing privacy-preserving methods mainly focus on post hoc suppression of generated node representations, while insufficiently considering how to directly weaken sensitive-attribute cues and their inferable associations during the representation recovery stage. To address this issue, this paper proposes a conditional diffusion-based graph privacy protection model with task-aware attention and hierarchical adversarial constraints, termed Task-Aware Adversarial Conditional Diffusion (TA-ACD). Specifically, a Graph Convolutional Network (GCN) is employed to construct a conditional diffusion denoising network for node representation recovery. Sensitive-attribute condition switching is introduced to alter the recovery path, thereby weakening the inferable association between sensitive attributes and node representations. Meanwhile, task-aware attention is embedded into the denoising process to preserve task-relevant discriminative information. Furthermore, node-level and global statistical adversarial constraints are designed to suppress residual sensitive information from both local and global perspectives. Finally, an alternating training strategy is adopted to achieve collaborative optimization between privacy protection and utility preservation. Experimental results on Yale, Rochester, and the Credit defaulter datasets show that TA-ACD achieves a competitive privacy–utility trade-off compared with representative baselines. On the Yale dataset, TA-ACD reduces sensitive-attribute inference accuracy from 86.3% to 62.1%, corresponding to a decrease of 24.2 percentage points, while maintaining a utility-attribute classification accuracy of 85.6%.

  • Personalized Privacy-Preserving Graph Neural Network Based on Community Awareness

    by Shaobo Du, Jinchuan Tang, Shuping Dang

    Existing differentially private graph neural network (GNN) methods predominantly rely on a uniform privacy budget and coarse-grained perturbation mechanisms, implicitly assuming homogeneous privacy sensitivity across graph nodes. However, real-world graphs exhibit pronounced community structures and topological heterogeneity, causing uniform privacy protection to simultaneously underprotect structurally critical regions and overperturb peripheral ones, thereby degrading both privacy guarantees and model utility. To address this limitation, we propose a community-aware personalized differential privacy (DP) framework for GNNs, which enables fine-grained node-level privacy protection by explicitly leveraging community structure. Specifically, we apply the Louvain algorithm to detect community structures and introduce a community importance evaluation method that integrates structural connectivity with node attribute similarity. Based on these importance scores, we design a personalized privacy budget allocation scheme that adaptively assigns differentiated privacy budgets to nodes under a global $(\epsilon,\delta)$ -DP constraint. We theoretically prove that the proposed mechanism satisfies node-level $(\epsilon,\delta)$ -DP and further demonstrate that it significantly enhances robustness against practical privacy threats such as attribute inference and link reconstruction attacks. Extensive experiments on five real-world graph datasets show that our approach consistently outperforms state-of-the-art DP-GNN baselines across multiple architectures, including graph convolutional network (GCN), graph isomorphism network (GIN), and Sample and aggregate (GraphSAGE), achieving a superior privacy–utility trade-off.

  • Open Access

    Dynamic Node Privacy Feature Decoupling Graph Autoencoder Based on Attention Mechanism

    by Yikai Huang, Jinchuan Tang, Shuping Dang

    Graph autoencoders’ inherent capability to capture node feature correlations poses significant privacy risks through attackers inference. Previous feature decoupling approaches predominantly apply uniform privacy protection across nodes, disregarding the varying sensitivity levels inherent in graph structures. To solve the above problems, we propose a novel dual-path graph autoencoder incorporating attention-aware privacy adaptation. Firstly, we design an attention-driven metric learning framework to quantify node-specific privacy importance through attention weights and select important nodes to construct the privacy distribution, so that realizing the dynamically privacy decoupling and reducing utility loss. Then, we introduce Hilbert-Schmidt Independence Criterion (HSIC) to measure the dependence between privacy and non-privacy information, which avoids the deviations that occur when using approximate methods such as variational inference. Finally, we use the method of alternating training to comprehensively evaluate the privacy importance of nodes. Experimental results on three real-world datasets—Yale, Rochester, and Credit Defaulter—demonstrate that our proposed method significantly outperforms existing approaches like PVGAE, GAE-MI, and APGE, where the inference accuracy regarding privacy decreased by 25.5%, but the accuracy rate of link prediction achieved the highest 84.7% compared to other methods.

  • Trust-Based Community Sharing and Leakage Tradeoff in Online Social Networks

    by Xinxin Li, Jinchuan Tang, Shuping Dang, Gaojie Chen

    In the online social networks (OSNs) and Social Internet of Things (SIOT), communities of interest (CoI) are often used to facilitate information sharing among user devices. However, the risk of information leaks across communities persists due to inadequate control over users’ sharing behavior. In this article, we propose a novel trust-based community sharing mechanism to control users who are contributing to high privacy leakage across communities of an OSN. In detail, we first formulate privacy loss in community based on the sensitivity and willingness of users in sharing. Second, we use this loss as the key determinant when updating trust to dynamically hold users accountable for privacy leakage. Third, we use an adjustable threshold to enable or disable sharing users and evaluate the amount of information shared before and after control, as well as the changes in community user trust. Finally, we propose an optimization method based on the upper confidence bound to make a tradeoff between information sharing and leakage through a payoff function over discretized thresholds. Simulations on three real OSNs datasets — BlogCatalog, Flickr, and YouTube — demonstrated that our proposed mechanism can effectively reduce community privacy loss by achieving the best payoff score of 1076.53, while the state-of-the-art baselines PDC-InfoSharing and user trust value scored 506.33 and 957.98, respectively.

  • Priority-Based Blockchain Packing for Dependent Industrial IoT Transactions

    by Chaofeng Lin, Jinchuan Tang, Shuping Dang, Gaojie Chen

    Blockchain plays a key role in establishing secure and decentralized Industrial Internet of Things (IIoT) systems. Currently, the dependent transactions generated by IIoT devices require a packing process to select a set of non-conflicted transactions, which results in significant delay and deviation of the transaction response time. In this paper, we propose a novel transaction packing algorithm named Priority-Pack to address the above issue. Firstly, we use directed acyclic graphs to model the dependent transactions in IIoT systems to establish the mathematical relationships between transaction priority and waiting time as well as dependencies. Secondly, we propose an algorithm to specify a higher priority to a transaction with longer waiting time without violating transaction dependencies. It eliminates the time required to traverse the subsets of transactions in other algorithms. Thirdly, to further reduce the response delay for transactions with the same priority level, we choose to first pack transactions with smaller sizes. We prove that this selection can achieve the lowest average response time. Finally, simulations are conducted to benchmark the Priority-Pack against the state-of-the-art algorithms including Fair-Pack and Random-Pack. The results demonstrate that Priority-Pack outperforms the others in terms of average response time and deviations.


  • Weight decay regularized adversarial training for attacking angle imbalance

    by Guorong Wang, Jinchuan Tang, Zehua Ding, Shuping Dang, Gaojie Chen

  • Membership inference attacks via spatial projection-based relative information loss in MLaaS

    by Zehua Ding, Youliang Tian, Guorong Wang, Jinbo Xiong, Jinchuan Tang, Jianfeng Ma