Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Xiaohan Zhang has a total of 14 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    MASteer: Multi-Agent Adaptive Steer Strategy for End-to-End LLM Trustworthiness Repair

    by Changqing Li, Tianlin Li, Xiaohan Zhang, Aishan Liu, Li Pan

    Large Language Models (LLMs) face persistent and evolving trustworthiness issues, motivating developers to seek automated and flexible repair methods that enable convenient deployment across diverse scenarios. Existing repair methods like supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) are costly and slow, while prompt engineering lacks robustness and scalability. Representation engineering, which steers model behavior by injecting targeted concept vectors during inference, offers a lightweight, training-free alternative. However, current approaches depend on manually crafted samples and fixed steering strategies, limiting automation and adaptability. To overcome these challenges, we propose MASteer, the first end-to-end framework for trustworthiness repair in LLMs based on representation engineering. MASteer integrates two core components: AutoTester, a multi-agent system that generates diverse, high-quality steer samples tailored to developer needs; and AutoRepairer, which constructs adaptive steering strategies with anchor vectors for automated, context-aware strategy selection during inference. Experiments on standard and customized trustworthiness tasks show MASteer consistently outperforms baselines, improving metrics by 15.36% on LLaMA-3.1-8B-Chat and 4.21% on Qwen-3-8B-Chat, while maintaining general model capabilities. MASteer demonstrates strong robustness, generalization, and practical value for scalable, efficient trustworthiness repair.

  • Open Access

    Multi-Scale Feature Fusion-Based Real-Time Anomaly Detection in Industrial Control Systems

    by Lin Xu, Kequan Shang, Xiaohan Zhang, Conghui Zheng, Li Pan

    Industrial control systems (ICSs) are a critical component of key infrastructure. However, as ICSs transition from isolated systems to modern networked environments, they face increasing security risks. Traditional anomaly detection methods struggle with complex ICS traffic due to their failure to fully utilize both low-frequency and high-frequency traffic information, and their poor performance in heterogeneous and non-stationary data environments. Moreover, fixed threshold methods lack adaptability and fail to respond in real time to dynamic changes in traffic, resulting in false positives and false negatives. To address these issues, this paper proposes a deep learning-based traffic anomaly detection algorithm. The algorithm employs the Hilbert–Huang Transform (HHT) to decompose traffic features and extract multi-frequency information. By integrating feature and temporal attention mechanisms, it enhances modeling capabilities and improves prediction accuracy. Additionally, the deep probabilistic estimation approach dynamically adjusts confidence intervals, enabling synchronized prediction and detection, which significantly enhances both real-time performance and accuracy. Experimental results demonstrate that our method outperforms existing baseline models in both prediction and anomaly detection performance on a real-world industrial control traffic dataset collected from an oilfield in China. The dataset consists of approximately 260,000 records covering Transmission Control Protocol/User Datagram Protocol (TCP/UDP) traffic between Remote Terminal Unit (RTU), Programmable Logic Controller (PLC), and Supervisory Control and Data Acquisition (SCADA) devices. This study has practical implications for improving the cybersecurity of ICSs and provides a theoretical foundation for the efficient management of industrial control networks.


  • FDFL: Fair and Discrepancy-Aware Incentive Mechanism for Federated Learning

    by Zhe Chen, Haiyan Zhang, Xinghua Li, Yinbin Miao, Xiaohan Zhang, Man Zhang, Siqi Ma, Robert H. Deng

    Federated Learning (FL) is an emerging distributed machine learning paradigm crucial for ensuring privacy-preserving learning. In FL, a fair incentive mechanism is indispensable for inspiring more clients to participate in FL training. Nevertheless, achieving a fair incentive mechanism in FL is an arduous endeavor, underscored by two significant challenges that persistently elude resolution within existing methodologies. Firstly, existing works overlook the issue of category distribution heterogeneity in contribution evaluation, leading to incomplete contribution evaluations. Secondly, the fact that malicious servers will dishonestly allocate rewards to save costs is not considered in existing work, which can be a barrier to client participation in FL. This paper introduces FDFL (Fair and Discrepancy-aware incentive mechanism for Federated Learning), a novel system addressing these concerns. FDFL encompasses two key elements: 1) Discrepancy-aware contribution evaluation approach; 2) Provable reward allocation approach. Extensive experiments on four model-dataset combinations demonstrate that, under the heterogeneous setting, our scheme improves accuracy by an average of 9.85% and 11.97% compared to FedAvg and FAIR, respectively.