Workshop Welcome Message
It is my great pleasure, on behalf of CRSC R&D institute, to welcome you to this invited session on AI-Empowered Safe and Secure Rail Control Systems, held in conjunction with ICRIT 2025.
The rapid development of next-generation ICT technologies, such as AI, Space-Air-Ground-Integrated Communications, 5G/6G, cloud computing, digital twin, is accelerating the transformation of various industrial sectors toward industry 4.0. Rail transportation, as a critical infrastructure, is no exception. In particular, enabled by AI, rail operation and control systems have significant promises to become more intelligent, resilient, safe, and secure, thereby contributing to highly efficient, sustainable, and low-carbon rail transportation. However, the integration of AI models into core rail operation and control systems presents significant challenges, due to the fact that these systems are inherently safety-critical, requiring stringent assurances of reliability, stability, and safety throughout their entire lifecycle. Therefore, the AI-driven rail applications must undergo rigorous processes of testing, validation, verification and certification to ensure compliance with functional safety and cybersecurity standards prior to real-world deployment.
This invited section aims to bring together the leading researchers and practitioners from academia, industry and government to share their insights, experiences, and lessons learned in the design, development, deployment, and operation of industrial control systems in the era of AI and digitalization. This session is dedicated to exploring both the challenges and opportunities in building autonomous, dependable, safe, and secure rail control systems by leveraging different AI techniques, such as LLMs, domain-specific foundation models, generative AI, deep learning, reinforcement learning, as well as classical machine learning algorithms. One hand, this special session seeks contributions on the application of AI techniques to enhance efficiency, automation, and decision-making capabilities in rail operation and control systems. On the other hand, it also aims to raise critical concerns and address essential questions related to the safety, trustworthiness, and cybersecurity risks of deploying AI models in such safety-critical systems. Through constructive discussions, this session will promote best practices, emerging norms, and standardization efforts to ensure that AI-empowered rail operation and control systems will achieve not only high performance, but also verifiable safety, security, and resilience.
I wish to thank all speakers, participants, and organizers for their contributions to this event, and I look forward to a successful session marked by productive discussions with a lasting impact.
Chair: Dr. Ling Liu, CRSC Co., Ltd.
Bio: Ling Liu is currently the Chief Engineer and Deputy General Manager of Beijing National Railway Research and Design Institute of Signal and Communication, China. He also serves as Director of the Engineering Research Center of the Railway Industry for Intelligent and Autonomous Train Control. He received his Ph.D. from Peking University, Beijing, China. Dr. Liu has led the development of multiple generations of the Chinese Train Control System (CTCS) and the European Train Control System (ETCS), which have been widely deployed in high-speed railways, intercity rail lines, and maglev systems. His research interests include safety-critical system technologies, train control system design, and intelligent transportation systems and applications.
Invited Session Speakers:
Professor Toshiaki Aoki, JAIST, Japan
Speaker Bio: Toshiaki Aoki is a professor of Graduate School of Advanced Science and Technology, Japan Advanced Institute of Science and Technology (JAIST). He focuses on the practical application of formal methods, particularly in the automotive domain. He has collaborated with automotive companies on joint projects and successfully applied formal methods to the verification of commercial products. His research interests include software engineering, formal methods, model checking, theorem proving, software testing, embedded systems, and automotive systems. He served ISORC 2004, ICFEM 2012, and TASE 2020 as program chairs as well as international conferences in the field of software engineering and formal methodsincluding ICSE/SEIP, APSEC, ISORC, PRDC, TASE, RTCSA, and FormaliSE as PCmembers. He is currently leading a JST/CREST project launched in 2023 titled “Formal Methods for Next Generation Automotive Systems”.
Title: Formal Methods for the Verification of AI in Automotive Systems
Abstract: We are working on practical applications of formal methods to automotive systems. Formal methods are a set of techniques to develop and verify systems based on mathematics and logic. They have been successfully applied to many safety-critical systems, including metro systems in Paris, France. We have also successfully applied formal methods to automotive operating systems such as OSEK/VDX, Classic AUTOSAR, and Adaptive AUTOSAR OSs. We would like to extend our target to more modern automotive system platforms, particularly those used in autonomous driving systems. Modern automotive systems consist of AI for perception and planning, control components, and basic software for high-performance computing. Recently, we launched a JST/CREST project titled Formal Methods and Verification Tools for Next-generation Automotive System Platforms, which focuses on these modern systems. This project aims to propose formal methods and verification tools to ensure the safety and reliability of next-generation automotive system platforms. In this talk, I will introduce the overview of the JST/CREST project after presenting the research results we have achieved so far.
PhD. Niu Fei,Prover Technology AB, Sweden
Speaker Bio: Fei Niu is the AI & Innovation Lead at Prover Technology AB, Sweden. He holds a Licentiate degree from KTH Royal Institute of Technology, Sweden, and a Master’s degree from Xi’an Jiaotong University, China. His early research explored software test automation by combining formal methods with machine learning. With over a decade of experience in formal methods and safety verification for rail control systems, he has led and contributed to projects with rail infrastructure managers and suppliers worldwide. At Prover, a global leader in formal verification and signaling design automation, he now drives AI innovation to advance process automation and safety-critical software engineering.
Prof. Daniel Fredholm
Speaker Bio:
Born: 1964
Phd in Mathematics, Stockholm University, 1996
PhD advisor: Per Martin-Löf
Title of dissertation: Intensional aspects of function definitions Works at Prover Technology since 2005, dealing with Formal Methods in the railway sector. Been heavily involved in developing the in-house methodology for Formal Verification and also in applying said methodology in numerous cases.
Title: How to use AI safely for rail control software – An industrial practice
Abstract:The integration of AI into safety-critical domains raises both opportunities and challenges. In this talk, we present industrial perspectives and practices on how AI can be applied safely for rail control software. We highlight how AI can improve the accessibility and usability of formal methods, supporting automation in requirements engineering, verification, and assurance processes. At the same time, we discuss considerations for the safety of AI-empowered rail control systems, drawing on practical insights and emerging approaches.
Professor Wei Wang, Xi’an Jiaotong University, China
Speaker Bio: Dr. Wei Wang is a full Professor and dean of school of software engineering, Xi’an Jiaotong University, China. He received the Ph.D. degree from Xi’an Jiaotong University, in 2006. He was a Post-Doctoral Researcher with the University of Trento, Italy, from 2005 to 2006. He was a Post-Doctoral Researcher with TELECOM Bretagne and with INRIA, France, from 2007 to 2008. He was also a European ERCIM Fellow with the Norwegian University of Science and Technology (NTNU), Norway, and with the Interdisciplinary Centre for Security, Reliability, and Trust (SnT), University of Luxembourg, from 2009 to 2011. He was a faculty member with Beijing Jiaotong University from 2011 to 2024. His recent research interests lie in privacy-preserving computation, AI security and blockchain. He has authored or co-authored over 100 peer-reviewed articles in various journals and international conferences, including IEEE TDSC, IEEE TIFS, IEEE TSE, USENIX Security, ACM CCS, AAAI, Ubicomp, IEEE INFOCOM. He received the ACM CCS 2023 Distinguished Paper Award. He is an Elsevier “highly cited Chinese Researchers”. He is an Associate Editor for IEEE Transactions on Dependable and Secure Computing, and an Editorial Board Member of Computers & Security and of Frontiers of Computer Science. He is a vice chair of ACM SIGSAC China.
Title: Data Security in the Construction of Intelligent Systems: Key Technologies and Typical Applications
Abstract: The construction of intelligent systems involves data, algorithms, and frameworks. Among these, data security and privacy protection constitute a critical component. However, with the rapid growth in data volume and the diversification of data modalities, data security and privacy protection face significant challenges throughout the system construction process. This talk will present the primary challenges in data security encountered during the development of intelligent systems, along with the corresponding key technologies, and introduce relevant typical applications.
Professor Jianwen Xiang, Wuhan University of Technology, China
Speaker Bio: Dr. Jianwen Xiang is a full professor of School of Computer and Artificial Intelligence, Wuhan University of Technology (WUT). He is currently the director of Network Information Center of WUT, director of Hubei International Scientific and Technological Cooperation Base of Highly Dependable Software Technology, and vice director of Engineering Research Center of Transportation Information and Safety (ERCTIS), Ministry of Education of China. He received the Ph.D. degree from Japan Advanced Institute of Science and Technology (JAIST) in 2005. He was a Post-Doctoral Researcher with JAIST from 2025 to 2027. He was a special researcher at National Institute of Advanced Industrial Science and Technology (AIST), Japan, from 2007 to 2008, and was an assistant manager at the Center Research Labs of NEC, Japan, from 2008 to 2014. His research Interests includes dependable computing, software engineering, and artificial intelligence. He has published over 100 peer-reviewed articles in various journals and international conferences, including IEEE TDSC, IEEE TSE, IEEE TIFS, IEEE TR, and RESS. He is a member of the Steering Committee of International Symposium on Software Reliability Engineering (ISSRE), and serves for a number of international conferences as general chair, PC chair, or PC/OC member, including ISSRE, DSN and WoSAR.
Title: Dual-View Spatio-Temporal Representation Learning for Intrusion Detection in Industrial Control System
Abstract:Industrial Control Systems (ICS) are essential for critical infrastructure, but their increasing integration with IT networks has expanded attack surfaces and heightened vulnerability to cyber threats. Most deep learning-based Intrusion Detection Systems (IDS) rely on the Traditional Spatio-Temporal Feature (TSTF) view, which often fails to capture the fine-grained unique characteristics of industrial protocols and ICS communication patterns, leading to higher false negatives and positives. To address this limitation, we propose a Segment-Based Spatio-Temporal Feature (SSTF) view that models temporal dependencies among same segments in different packets and spatial correlations between different segments in ICS traffic. We further introduce DVSTdetector, a dual-view IDS framework that combines both TSTF and SSTF views and perform spatio-temporal representation learning from global and local perspectives in parallel. Experiments on public (WDT) and private (XLP) ICS datasets validate the effectiveness of DVSTdetector, demonstrating its superior robustness and detection performance compared to existing state-of-the-art methods.
Professor Zonghua Zhang, CRSC Co., Ltd.
Speaker Bio: Zonghua is now working on cyber resilience for future rail transit systems. Previously, he served as Chief Expert at Huawei Paris for four and a half years, where he led research on resilient and trustworthy AI for Autonomous Driving Networks. Before diving into the industry, Zonghua has spent more than 15 years in academia at different institutions (Professor at IMT, Researcher at NICT, INRIA, JAIST, University of Waterloo). He holds an HDR diploma (UPMC, France) in computer science, and a Ph.D. Degree (JAIST, Japan) in information science. He has been actively working at the intersections of cyber resilience, networking, and applied machine learning. Zonghua has contributed, as either PI or key contributor, to more than a dozen national and international research projects, with the topics ranging from anomaly detection and root cause analysis to trust management and software-defined cyber defense. These research projects have led to the publication of 100+ research articles in well-recognized international journals and conferences. His accolades include Thailand’s NRCT Excellent Ph.D. Supervisor Award, JSPS Invitational Fellowship, multiple Best Paper Awards, and Huawei’s President Award & Technological Breakthrough Awards. He co-founded and chaired the international Workshop ARTMAN (Resilient and Trustworthy MAchine learning-driveN systems) on three occasions (twice co-located with ACM CCS and once with ACSAC) and has served on the editorial board of journal Computers & Security for over a decade.
Title: The Paradigm Shift of Cyber Resilience in Digitalized Train Control Systems
Abstract: The rapid development of ICT, from AI and cloud computing to 5G/6G, is driving a profound digital transformation in train control systems. These advances promise to cut system complexity and operational costs, while boosting safety and efficiency. Yet, as the cyber dimension of rail transit expands, security remains paramount: even a small cyber breach could cause major safety consequences in the physical world. In this talk, I will discuss the unique requirements and emerging challenges of making future train control systems as resilient as, or even more resilient than, today’s proven designs. I will also outline the potential needs for solid theoretical foundations, practical technical approaches, relevant standards, and real-world implementations. The talk will conclude with a proposed cyber resilience paradigm aimed at meeting the demands of the rail transit control systems in the digital age.
