Frontiers of DE & AI @ IEEE ICDE 2025

---

Abstract

In this presentation, I'll talk about how important location data and its analysis are for a number of different application domains, such as urban planning, transportation, public health, and security. But, there's a problem: this data is sensitive, especially with all the growing privacy concerns, potentially hindering its accessibility. So, I'll discuss different ways to keep location data private while still making it useful for these applications.
I'll focus on deep neural network methods we're exploring to 1) release aggregated location data and 2) create synthetic location visit sequences. Finally, I will conclude by reviewing some emerging applications on the horizon, accompanied by the associated challenges and open questions.

---

Biography

Cyrus Shahabi is a Professor of Computer Science, Electrical & Computer Engineering and Spatial Sciences; Helen N. and Emmett H. Jones Professor of Engineering; and the director of the Integrated Media Systems Center (IMSC) at USC’s Viterbi School of Engineering. He also served as USC's Thomas Lord Department of Computer Science from 2017 to 2022. He was co-founder of two startups, Geosemble Technologies and Tallygo, which both were acquired in July 2012 and March 2019, respectively. He received his B.S. in Computer Engineering from Sharif University of Technology in 1989 and then his M.S. and Ph.D. Degrees in Computer Science from the University of Southern California in May 1993 and August 1996, respectively. He authored two books and more than three hundred research papers in databases, GIS, and multimedia, and he has over 14 US patents.
Dr. Shahabi has received funding from several agencies such as NSF, NIJ, NASA, NIH, DARPA, AFRL, IARPA, NGA, and DHS, as well as several industries such as Chevron, Cisco, Google, HP, Intel, Microsoft, NCR, NGC, and Oracle. He chaired the founding nomination committee of ACM SIGSPATIAL for its first term (2008-2011 term) and served as the chair of ACM SIGSPATIAL for the 2017-2020 term. He was an Associate Editor of IEEE Transactions on Parallel and Distributed Systems (TPDS) from 2004 to 2009, IEEE Transactions on Knowledge and Data Engineering (TKDE) from 2010 to 2013, VLDB Journal from 2009 to 2015 and PVLDB (Vol. 16) in 2023. He is on the ACM Transactions on Spatial Algorithms and Systems (TSAS) editorial board and ACM Computers in Entertainment. He was the founding chair of the IEEE NetDB workshop and the general co-chair of SSTD’15, ACM GIS 2007, 2008, and 2009. He has been PC co-chair of several conferences, such as APWeb+WAIM’2017, BigComp’2016, MDM’2016, DASFAA 2015, IEEE MDM 2013, IEEE BigData 2013 and VLDB 2024. He regularly serves on the program committee of major conferences such as VLDB, SIGMOD, IEEE ICDE, ACM SIGKDD, IEEE ICDM, and ACM Multimedia.
Dr. Shahabi is a fellow of IEEE and NAI (National Academy of Inventors). He received the ACM Distinguished Scientist Award 2009, the 2003 U.S. Presidential Early Career Awards for Scientists and Engineers (PECASE), the NSF CAREER award in 2002, and the 2001 Okawa Foundation Research Award. He received the ACM SIGSPATIAL 2023 10-Year Impact Award in 2023. He was also a recipient of the US Vietnam Education Foundation (VEF) faculty fellowship award in 2011 and 2012, an organizer of the 2011 National Academy of Engineering “Japan-America Frontiers of Engineering” program, an invited speaker in the 2010 National Research Council (of the National Academies) Committee on New Research Directions for the National Geospatial-Intelligence Agency, and a participant in the 2005 National Academy of Engineering “Frontiers of Engineering” program.

---

Abstract

Recent breakthroughs in deep learning, particularly embedding models, have revolutionized how we represent diverse data types—from images to text chunks—by transforming them into high-dimensional vectors that capture semantic information. Consequently, k-approximate nearest neighbor (k-ANN) search in high-dimensional spaces has become a critical operation across domains like information retrieval, recommendation systems, and large language models. In addition, the rise of retrieval-augmented generation (RAG) has further amplified the importance of efficient k-ANN search, as it serves as a core component for grounding LLMs in relevant data. In this talk, we explore new techniques to enhance k-ANN performance for large-scale, high-dimensional vector databases, addressing scalability, accuracy, and speed challenges in modern applications.

---

Biography

Jeffrey Xu Yu is a Professor in the Department of Systems Engineering and Engineering Management at The Chinese University of Hong Kong. His primary research interests include graph algorithms, graph processing systems, graph neural networks, and query processing in database systems. Dr. Yu has held several prestigious roles, including Information Director and member of the ACM SIGMOD Executive Committee (2007–2011), Associate Editor of IEEE Transactions on Knowledge and Data Engineering (TKDE) (2004–2008), and Associate Editor of the VLDB Journal (2007–2013). He currently serves as an Associate Editor for ACM Transactions on Database Systems (TODS) and the WWW Journal, among others. Dr. Yu has been actively involved in the organization and program committees of numerous international conferences and workshops. He has served as PC Co-Chair for conferences such as APWeb'04, WAIM'06, APWeb/WAIM'07, WISE'09, PAKDD'10, DASFAA'11, ICDM'12, NDBC'13, ADMA'14, CIKM'15, Bigcomp'17, DSAA'19, CIKM'19, and DASFAA'20, and as Conference General Co-Chair for APWeb'13, ICDM'18, and ADC'22.

---

Abstract

As artificial intelligence (AI) and large language models (LLMs) increasingly influence every facet of our lives, ensuring the privacy of user data has become paramount. In this talk, I will review the common privacy attacks for inferring training data from a trained AI model and common defenses for building privacy-enhanced models using privacy sensitive data. I will then present our recent works and discuss open challenges related to: 1) new privacy attacks under the fine-tuning paradigm using pre-trained LLMs, and 2) end-to-end privacy defenses across the life cycle of AI and LLMs.

---

Biography

Li Xiong is a Samuel Candler Professor of Computer Science and Biomedical Informatics at Emory University. She has a Ph.D. from Georgia Institute of Technology, an MS from Johns Hopkins University, and a BS from the University of Science and Technology of China. Her research lab, Assured Information Management and Sharing (AIMS), conducts research on trustworthy and privacy-enhancing data-driven AI solutions for healthcare, public health, and spatial intelligence. She is recognized as an IEEE fellow (2022) and AAAS fellow (2024) for her contributions on privacy-preserving and secure data sharing and analytics. She has published over 200 papers and received seven best paper or runner up awards. Her research has been supported by both governments (NSF, NIH, IARPA, AFOSR) and industry/foundations (Mistubishi, Cisco, AT&T, Google, IBM). More details are at http://www. cs.emory.edu/~lxiong.

---

Abstract

With the convergence of data and AI, the database community has actively explored two key directions: leveraging AI to tackle challenges in database systems (AI4DB) and developing data systems to support AI applications. This talk will first cover AI4DB, focusing on the following topics: 1) Query Plan Representation as a foundation of AI4DB, serving as input for various AI4DB tasks. 2) Paradigms of AI4DB – We discuss “Replacement,’ aiming to learn machine learning (ML) models to replace traditional components (e.g., learned indexes), and “Enhancement,” aiming to apply ML to improve existing database components (e.g., optimizing indexes without replacement). 3) LLMs for Query Rewriting – A brief introduction to using large language models (LLMs) in query optimization.

---

Biography

Gao Cong is currently a Professor in the College of Computing and Data Science at Nanyang Technological University (NTU), and is the head of Division of Data Science of NTU. He previously worked at Aalborg University, Microsoft Research Asia, and the University of Edinburgh. His current research interests include AI4DB, spatial data management, spatial-temporal data mining, and recommendation systems. His citation in Google Scholar was over 20,000 with H-index 73. He received the best paper runner-up awards at the WSDM'20 and WSDM'22 conferences for two of his research papers. He is an associate editor for ACM Transactions on Database Systems (TODS), IEEE TKDE, and the vice Chair of ACM KDD Singapore chapter.

---

Abstract

Recently, "pre-training and fine-tuning" has been adopted as a standard workflow for many graph tasks since it can take general graph knowledge to relieve the lack of graph annotations from each application. However, graph tasks with node level, edge level, and graph level are far diversified, making the pre-training pretext often incompatible with these multiple tasks. This gap may even cause a negative transfer to the specific application, leading to poor results. Inspired by the prompt learning in natural language processing (NLP), which has presented significant effectiveness in leveraging prior knowledge for various NLP tasks, in the first work, we study the prompting topic for graphs with the motivation of filling the gap between pre-trained models and various graph tasks. We propose a novel multi-task prompting method for graph models. To narrow the gap between various graph tasks and state-of-the-art pre-training strategies, we study the task space of various graph applications and reformulate downstream problems to the graph-level task. Afterward, we introduce meta-learning to efficiently learn a better initialization for the multi-task prompt of graphs so that our prompting framework can be more reliable and general for different tasks. In the second work, we study cross-domain graph pre-training and propose a novel approach called Graph COordinators for PrEtraining (GCOPE) that harnesses the underlying commonalities across diverse graph datasets to enhance few-shot learning. By successfully leveraging the synergistic potential of multiple graph datasets for pretraining, our work stands as a pioneering contribution to the realm of graph foundational model.

---

Biography

Hong Cheng is a Professor in the Department of Systems Engineering and Engineering Management, The Chinese University of Hong Kong. She received the Ph.D. degree from the University of Illinois at Urbana-Champaign in 2008. Her research interests include data mining, database systems, and machine learning. She received best research paper award at SIGKDD'23. She received the 2010 Vice-Chancellor's Exemplary Teaching Award at The Chinese University of Hong Kong.

---

Abstract

Imagine a scenario where we could communicate with a relational query optimizer through messaging platforms like WeChat or WhatsApp, asking questions related to query processing and optimization. If realized, such a tool could prove invaluable in database administration and education, among others. In this talk, we share our ongoing efforts to develop a user-friendly tool, ChatQPT, which facilitates chat-based interactions with a relational query engine (PostgreSQL). We will specifically highlight the complex challenges encountered in addressing this problem and our initial approaches to overcoming them.

---

Biography

Sourav S. Bhowmick is an Associate Professor in the College of Computing & Data Science (CCDS), Nanyang Technological University, Singapore. His research interests are in data management, human-data interaction, and data analytics. His research has appeared in premium venues such as ACM SIGMOD, VLDB, and VLDB Jour nal. He is co-recipient of Best Paper Awards in ACM CIKM 2004, ACM BCB 2011, VLDB 2021, and ER 2023. Sourav is serving as a member of the SIGMOD Executive Committee, SIGMOD Awards Committee, a regular member of the PVLDB advisory board, and an elected trustee of the VLDB Endowment. He is a co-recipient of several service awards including VLDB Service Award in 2018, Distinguished AE Award in SIGMOD 2021, SIGMOD 2023 and VLDB 2022, and Distinguished Reviewer Award in VLDB 2020 and VLDB 2023. He was inducted into Distinguished Members of the ACM in 2020. Sourav is a strong advocate of research that directly or indirectly impacts end users.

---

Abstract

Graphs are increasingly used in real-world applications involving human users, yet they often suffer from inconsistencies that violate domain-specific constraints. Repairing graph errors typically requires expert knowledge and might lead to possible conflicts among violations. Meanwhile, generative AI has demonstrated the ability to produce commonsense knowledge, which can be valuable for graph correction.

In this talk, I will present our work on generating high-quality graphs through a user-centric approach and a graph constraint mining method using large language models. Their interplay and integration highlights promising directions for future research around data engineering and AI .

---

Biography

Angela Bonifati is a Distinguished Professor of Computer Science at Lyon 1 University and at the CNRS Liris research lab, where she leads the Database Group. She is also an Adjunct Professor at the University of Waterloo in Canada from 2020 and a Senior member of the French University Institute (IUF) from 2023. Her current research interests are on several aspects of data management, including graph databases, knowledge graphs and data integration with additional interests in data science and AI. She has co-authored more than 200 publications in top venues of the data management field, including five Best Paper Awards, two books and an invited paper in ACM Sigmod Record 2018. She is the recipient of the prestigious IEEE TCDE Impact Award 2023 and a co-recipient of an ACM Research Highlights Award 2023. She is the Program Chair of IEEE ICDE 2025, the General Chair of VLDB 2026 and an Associate Editor for the Proceedings of VLDB and for several other journals, including the VLDB Journal, IEEE TKDE and ACM TODS. She was the Program Chair of ACM Sigmod 2022 and EDBT 2020. She is the Chair of the EDBT Executive Board (2020-2025), a member of the PVLDB Board of Trustees (2024-2029) and a member of the IEEE TCDE executive committee (2024-2029).

---

Abstract

In recent years, differential privacy (DP) has become a well-accepted standard for privacy protection, and AI models have been immensely successful in practical applications. Combining the two is a promising idea, since it could enable the release of accurate models built on sensitive data such as medical records. In this talk, we will revisit the existing techniques for training AI models with DP, and present improved solutions based on importance sampling and zeroth order optimizations.

---

Biography

Xiaokui Xiao is a professor at the School of Computing, National University of Singapore. His research focuses on data management and analytics, especially on data privacy and algorithms for large data. He was a recipient of the 2024 ACM SIGMOD Test-of-Time Award, 2022 ACM SIGMOD Research Highlight Award, and VLDB 2021 Best Research Paper Award. He is an IEEE fellow, ACM distinguished member, and trustee of the VLDB Endowment.

---

Abstract

Graphs are a natural way to model relationships among entities, and graph computing provides the tools to analyze these relationships at scale. As data becomes more dynamic, stream computing enables real-time processing of continuous data streams, capturing timely insights from ever-changing environments. When combined, graph streams offer a dynamic representation of evolving structures—tracking how entities and their interactions change over time. This fusion of graph and stream computing lays the groundwork for the next generation of big models, which must learn not just from large volumes of data, but from data that is richly structured and constantly evolving.

---

Biography

Xike Xie is currently a research professor at the Suzhou Institute for Advanced Research, University of Science and Technology of China. Previously, he served as an Assistant Professor at Aalborg University in Denmark. His research primarily focuses on database systems, data mining, and data-driven machine learning. His recent interests include graph computing, stream processing, data exploration, and large language models (LLMs). He has authored over 100 academic papers, with more than 50 published in CCF A conferences and journals. He has led three projects funded by the National Natural Science Foundation of China (NSFC) and has served on the program committees of over 30 international conferences, as well as a reviewer for more than 20 international journals.

---

Abstract

Differential privacy (DP) has emerged as a powerful paradigm for protecting individual information in graph data, enabling the release of meaningful statistics without revealing sensitive relationships. However, applying DP to graph analytics—especially under strong privacy notions like node-level DP—introduces severe utility challenges, particularly in decentralized settings where no trusted server is available. High sensitivity of graph queries and overlapping data across parties often result in excessive noise, rendering naive local or federated approaches ineffective.

This talk presents a unified view of recent advances that combine differential privacy with cryptographic techniques to overcome these challenges. We explore four systems that illustrate how different pieces of the puzzle fit together: local graph projection with secure parameter tuning, federated analytics with private set union protocols, private synthetic graph generation benchmarking, and triangle counting via secret sharing and distributed noise. These approaches demonstrate how cryptographic primitives—such as order-preserving encryption, secure aggregation, and multi-party computation—can significantly improve the utility of privacy-preserving graph analytics while reducing trust assumptions. The talk aims to provide intuitions, insights, and open questions to guide future research at the intersection of DP and cryptography in graph settings.

---

Biography

Yang Cao is an Associate Professor at the Department of Computer Science, Institute of Science Tokyo (Science Tokyo, formerly Tokyo Tech), and directs the Trustworthy Data Science and AI (TDSAI) Lab. He is passionate about studying and teaching on algorithmic trustworthiness in data science and AI. Two of his papers on data privacy were selected as best paper finalists in top-tier conferences IEEE ICDE 2017 and ICME 2020. He received the IEEE Computer Society Japan Chapter Young Author Award 2019, Database Society of Japan Kambayashi Young Researcher Award 2021. His research projects were/are supported by JSPS, JST, MSRA, KDDI, LINE, WeBank, etc.

---

Abstract

High-dimensional vector data lies at the core of numerous modern applications, from recommendation systems to large-scale retrieval and retrieval-augmented generation. Effectively managing and processing such data presents both significant opportunities and challenges. A key enabler in this context is vector quantization, which compresses high-dimensional vectors while preserving essential similarities. In this talk, I will begin by exploring why vector quantization is crucial for sclable and efficient high-dimensional vector management. I will then present the general framework of vector quantization and discuss existing popular schemes. Building on this foundation, I will introduce RaBitQ, a recent advance that provides optimized approaches for binary and scalar quantization and achieves asymptotic optimality. The talk will conclude with a discussion on future research directions of vector quantization.

---

Biography

Cheng LONG is an Associate Professor at the College of Computing and Data Science (CCDS), Nanyang Technological University (NTU). He earned his Ph.D. degree from Hong Kong University of Science and Technology (HKUST) in 2015. He has research interests broadly in data management and data mining. More recently, he has been working in high-dimensional vector data management (and its applications in large models such as retrieval-augmented generative AI). His work has garnered recognition and accolades, including the prestigious "Best Research Award" from ACM-Hong Kong, the "Fulbright-RGC Research Award" granted by the Research Grant Council (Hong Kong), the "PG Paper Contest Award" bestowed by IEEE-HK, and the "Overseas Research Award" received from HKUST.

---

Abstract

Since its inception, the database system has served as a cornerstone of information systems for over half a century, undoubtedly standing the test of time. However, the recent surge in AI, particularly in machine learning powered applications, poses new challenges for underlying data infrastructures. In this talk, I will argue that while it may seem that traditional database systems are becoming inadequate or even irrelevant for these AI-powered applications, the fundamental principles of databases still play a vital role in supporting these emerging applications. We will delve into concrete examples, showing how primary keys, integrity constraints, and database queries can assist in interpreting machine learning predictions, with just a dash of imagination. Moreover, we will also explore the new challenges and opportunities these AI applications present to our understanding of the age-old concept of database isolation.

---

Biography

Yang Cao is an Assistant Professor (Lecturer) at the University of Edinburgh. His research spans database systems, data management, data quality and the intersection of databases and machine learning, with a focus recently on the interoperability across data models. He is a recipient of the Royal Academy of Engineering Research Fellowship, the SIGMOD Research Highlight Award, and the SIGMOD Best Paper Award. His work has been invited to the ACM TODS special issues for "Best-of-PODS" (for theory), "Best-of-SIGMOD" (for systems), and the "Best-of-BICOD" Computer Journal special issue (for applications).

---

Abstract

Local differential privacy (LDP) has emerged as the de facto standard for privacy preservation in big data analytics and deep learning. However, real-world adoption still faces critical challenges. In this talk, I will discuss recent advances in making LDP more practical and robust. First, I will address the limitations of the semi-honest threat model and present novel defenses against poisoning attacks in LDP systems. Next, I will explore how LDP can be efficiently integrated into multi-service architectures and applied to multi-class data mining. Finally, I will highlight key open challenges for the next generation of privacy-aware deep learning and large-scale analytics.

---

Biography

Haibo Hu is a professor with the Department of Electrical and Electronic Engineering, The Hong Kong Polytechnic University. His research interests include cybersecurity, data privacy, and adversarial machine learning. He has published over 180 research papers in refereed journals, international conferences, and book chapters, and is granted 6 US patents and 4 China/HK patents. He is the recipient of a number of titles and awards, including IWAIT 2021 Best Paper Award, IEEE MDM 2019 Best Paper Award, WAIM Distinguished Young Lecturer, ACM-HK Best PhD Paper, Microsoft Imagine Cup, and GS1 Internet of Things Award. He is a senior member of ACM, IEEE and CCF, and a certified Cisco CCNA Security Trainer.