| Keynote Speaker Ⅰ |
Prof. Hongwei Li
University of Electronic Science and Technology of China, China
IEEE Fellow
Biography: Hongwei Li is the Director of the Center for Faculty Development at the University of Electronic Science and Technology of China. He is an IEEE Fellow (2024), a Distinguished Professor of the National Major Talent Project (2019), a member of the expert group for a major special project under the Ministry of Science and Technology's (MOST) 15th Five-Year Plan, and the group leader for the "Cyberspace Data Governance" direction of the MOST 14th Five-Year Plan key special project on "Cyberspace Security Governance." Additionally, he serves as the Director of the National Discipline Innovation and Talent Introduction Base for Data Security Governance, Chair of the Communication and Information Security Technical Committee under the IEEE Communications Society, and an IEEE Vehicular Technology Society Distinguished Lecturer.
His research interests focus on data security and artificial intelligence security. He has published 85 papers in JCR-1 journals and CCF-A conferences. He is the recipient of several prestigious awards, including the 2025 IEEE Signal Processing Society Best Paper Award, the ACM CCS 2025 Distinguished Paper Award, the First Prize of the 2019 National Science and Technology Progress Award, the First Prize of the 2021 Wu Wenjun AI Science and Technology Progress Award, and the Second Prize of the 2021 Natural Science Award from the Chinese Institute of Electronics.
Speech Title: The Current State and Future Prospects of Agent Security
Abstract: Agents are currently experiencing widespread adoption. However, numerous security challenges remain during their practical application. This talk will analyze the current security landscape of agents from multiple dimensions, including their practical significance, existing challenges, and our team's research foundation, while also outlining future development trends.
| Keynote Speaker Ⅱ |
Prof. Xinwang Liu
National University of Defense Technology, China
Biography: Xinwang Liu is a Professor and Doctoral Supervisor at the College of Computer Science, National University of Defense Technology, and a recipient of the National Science Fund for Distinguished Young Scholars (2023) and the Excellent Young Scientists Fund (2019). He serves as the Principal Investigator for Key Projects of the NSFC and the Sci-Tech Innovation 2030 Major Project, and is a core member of the NSFC Innovative Research Group. Focusing on machine learning and data mining, Professor Liu has published over 200 papers in CCF-A journals and conferences, including 30 in IEEE T-PAMI (with 3 as the sole author), and has garnered over 20,000 Google Scholar citations, ranking among the World's Top 2% Scientists for three consecutive years (2022–2024). His extensive research achievements have earned him numerous prestigious accolades, including the First Prizes of the CCF Natural Science Award (2025, ranked 1st), Wu Wenjun AI Natural Science Award (2024, 1st), and Beijing Science and Technology Progress Award (2024, 2nd), along with multiple Hunan Provincial and CSIG Natural Science Awards. Furthermore, he serves as an Associate Editor for leading international journals such as IEEE T-KDE, IEEE T-NNLS, and IEEE T-CYB, and acts as an Area Chair for premier international conferences including ICML and NeurIPS.
Speech Title: Efficient Federated Incomplete Multi-View Clustering
Abstract: Multi-view clustering (MVC) leverages complementary information from diverse data sources to enhance clustering performance. However, its practical deployment in distributed and privacy-sensitive scenarios remains challenging. Federated multi-view clustering (FMVC) has emerged as a potential solution, but existing approaches suffer from substantial limitations, including excessive communication overhead, insufficient privacy protection, and inadequate handling of missing views. To address these issues, we propose Efficient Federated Incomplete Multi-View Clustering (EFIMVC), a novel framework that introduces a localized optimization strategy to significantly reduce communication costs while ensuring theoretical convergence. EFIMVC employs both view-specific and shared anchor graphs as communication variables, thereby enhancing privacy by avoiding the transmission of sensitive embeddings. Moreover, EFIMVC seamlessly extends to scenarios with missing views, making it a practical and scalable solution for real-world applications. Extensive experiments on benchmark datasets demonstrate the superiority of EFIMVC in clustering accuracy, communication efficiency, and privacy preservation. Our code is publicly available at \href{https://github.com/Tracesource/EFIMVC}{https://github.com/Tracesource/EFIMVC}.
| Keynote Speaker Ⅲ |
Prof. Xiang Bai
Huazhong University of Science and Technology, China
IEEE Fellow, IAPR Fellow
Biography: Xiang Bai, Professor and Doctoral Supervisor at Huazhong University of Science and Technology, recipient of the National Science Fund for Distinguished Young Scholars, IEEE/IAPR Fellow, and Associate Editor-in-Chief (A-EIC) of the international journal Pattern Recognition. His research primarily focuses on computer vision, pattern recognition, and multimodal foundation models. He has published over 200 papers in top-tier international journals and conferences such as Nature Machine Intelligence, IEEE TPAMI, and CVPR. He serves as an Associate Editor of IEEE TPAMI (a leading international journal), Area Chair for top conferences including CVPR, ICCV, ECCV, AAAI, IJCAI, and NeurIPS, and General Chair of ICDAR 2025 (International Conference on Document Analysis and Recognition). His honors include the ACL 2024 Best Paper Award, the 2024 Hubei Youth Science and Technology Innovation Award, the First Prize of Hubei Natural Science Award (ranked 1st) in 2023, the First Prize of Natural Science Award of the China Society of Image and Graphics (ranked 1st) in 2021, and the IAPR/ICDAR Young Investigator Award in 2019. He is currently a Standing Director of the China Society of Image and Graphics and Director of its Youth Committee.
Speech Title: From 3D World Cognition to Complex Physical Execution
Abstract: This report aims to explore how to construct embodied intelligent systems capable of transitioning from environmental understanding to complex physical execution. The discussion will center on three core aspects spanning from 3D world cognition to closed-loop decision-making. First, we investigate how to build a unified world model that enables collaborative reasoning between semantic understanding of 3D scenes and future geometric evolution prediction, while also mining implicit 3D spatial knowledge within generative models, thereby establishing a reliable perceptual foundation of the physical world for intelligent agents. Second, at the level of dynamic decision-making, focusing on autonomous driving scenarios, we analyze how to bridge the gap between high-level semantic reasoning of vision-language models and low-level continuous actions, and enhance the closed-loop control and decision-making capabilities of vehicles in complex interactive environments through instruction-driven action generation and online trial-and-error learning. Finally, we extend the scope to generalizable robotics, exploring how intelligent agents can develop stable spatiotemporal perception and precise manipulation capabilities in dynamic mobile environments, as well as how to improve overall efficiency and coordination during the execution of complex tasks. These interconnected explorations collectively outline a development pathway for embodied agents from spatial cognition and scene reasoning to physical manipulation.
| Keynote Speaker Ⅳ |
Prof. Chao Shen
Xi’an Jiaotong University, China
IEEE Fellow
Biography: Chao Shen, IEEE Fellow, Chair Professor of Xi’an Jiaotong University, Cheung Kong Scholar Professor, Head of the National Natural Science Foundation of China Innovation Group, Head of the Ministry of Education Innovation Team, Director of the Ministry of Education Key Laboratory, Recipient of the Science Exploration Award, DAMO Academy Young Orange Award, MIT TR35 China, and IEEE SMC Early Career Award, Chief Scientist of the National Key Research and Development Program of China, and Chair of the IEEE Trustworthy and Controllable Intelligent Systems Technical Committee.
Speech Title: Data-Intelligence Chain Security: From Small Models to Large Models to Embodied Intelligence
Abstract: Data-driven intelligent systems, through the integration of advanced sensing methods, control algorithms, decision-making technologies, and artificial intelligence, have evolved from small models to large models to embodied intelligence. As these systems have become increasingly sophisticated, their security challenges have expanded beyond traditional network and system levels to encompass the entire chain of data, models, algorithms, and applications. Ensuring the security of artificial intelligence has become a core challenge in the design, development, and deployment of intelligent systems. This report systematically reviews research progress in the field of artificial intelligence security. Focusing on security dimensions such as confidentiality, integrity, and privacy, and starting from the component structure of the intelligent supply chain, it introduces the team's research and practices in the areas of intrinsic and extrinsic security within the intelligent supply chain.
| Keynote Speaker Ⅴ |
Prof. Bin Xiao
The Hong Kong Polytechnic University, China
IEEE Fellow
Biography: Dr. Bin Xiao is a professor at the Department of Computing, the Hong Kong Polytechnic University, Hong Kong. Prof. Xiao received the B.Sc and M.Sc degrees in Electronics Engineering from Fudan University, China, and a Ph.D. degree in computer science from the University of Texas at Dallas, USA. His research interests include AI security, data privacy, Web3, and blockchain systems.
He is currently an Associate Editor of the IEEE Transactions on Information Forensics and Security (TIFS) and IEEE Transactions on Cloud Computing (TCC). He has been the associate editor of the IEEE Internet of Things Journal, IEEE Transactions on Network Science and Engineering, and Elsevier Journal of Parallel and Distributed Computing. He is the IEEE Fellow, and has been the IEEE ComSoc Distinguished Lecturer, and the chair of the IEEE ComSoc CISTC committee from 2024 to 2025. He has been the program co-chair of IEEE CNS2025, track co-chair of IEEE ICDCS2026, ICDCS2022, the symposium track co-chair of IEEE Globecom 2024, ICC2020, ICC 2018, and Globecom 2017, and the general chair of IEEE SECON 2018.
Speech Title: Jailbreaking and Knowledge Poisoning Attacks on AI Agents
Abstract: Jailbreaking and Knowledge Poisoning Attacks on AI Agents
AI agents increasingly combine large language models (LLMs), multimodal perception, and retrieval-augmented generation (RAG) to support complex reasoning and decision-making. However, these modular architectures introduce new security risks beyond traditional model-level vulnerabilities. In this talk, we examine two emerging attack surfaces in AI agents: jailbreaking the agent’s reasoning core and poisoning its external knowledge memory.
First, we present video-driven jailbreaking attacks on multimodal LLMs, showing that safety alignment is weaker in the video modality. By constructing safety-proximal typographic videos with diverse frames, attackers can bypass safety defenses more effectively than image-based methods. Second, we analyze black-box knowledge poisoning attacks on retrieval-augmented diffusion models, where adversaries inject malicious embeddings into the knowledge base and jointly optimize triggers to manipulate both retrieval and generation. These works reveal that AI agents are vulnerable at the representation and memory levels, highlighting the need for principled defenses in multimodal and retrieval-augmented systems.