Professor Xiaokui Xiao and Collaborators Take Home Three Major Awards in AI, Data Systems, and Privacy

Three papers co-authored by Professor Xiaokui Xiao have won top honours at leading international conferences this year, recognising work spanning AI system efficiency, differential privacy auditing, and graph algorithms.

“I am deeply grateful to our students, research staff, former students and collaborators, whose hard work and creativity made these results possible, and to NUS Computing for its strong support of our research,” said Prof Xiao. 

ACM SIGKDD Best Student Paper Award
KDD 2026, 9–13 August, Jeju, South Korea 

Yiqian Huang, a Year 3 PhD candidate in Prof Xiao’s group, won the award for SCOPE: Cost-Efficient Model Selection for Compound AI Systems under Quality Constraints, co-authored with Shiqi Zhang, a postdoctoral researcher in Xiao's group, Tianyuan Jin, a former PhD student of Xiao’s, who is now an Assistant Professor in the Thrust of Data Science and Analytics (DSA) at the Hong Kong University of Science and Technology (Guangzhou), and Prof Xiao.

The paper proposes a way to choose a cost-effective combination of large language models for a multi-step AI system, while ensuring that the system as a whole meets a required quality level. In experiments, SCOPE met that quality level at as little as one-sixth of the cost of the strongest competing approach.

IEEE Symposium on Security and Privacy Distinguished Paper Award
IEEE S&P 2026, 18–21 May, San Francisco, USA 

Auditing Apple's DifferentialPrivacy.framework: Implementation Bugs, Misconfigurations, and Practical Risks put a widely trusted privacy system to the test. Apple's differential privacy framework runs on more than 2.35 billion devices, quietly reassuring users that the data their phones send back can't be traced to them individually. The team, led by Rishav Chourasia and Ergute Bao, both former PhD students of Xiao’s now with BetterData and as a postdoctoral researcher at Mohamed bin Zayed University of Artificial Intelligence respectively, together with Uzair Javaid, and Prof Xiao, found that several mechanisms did not deliver their stated privacy guarantees.

In particular, some secure-aggregation protocols had local privacy protection switched off entirely, and out on the open internet, the team found analytics logs that could be decoded to reveal what websites someone had visited, or which emojis they had typed. The team disclosed the vulnerabilities to Apple, and was informed that the affected mechanisms had been deprecated.

ACM PODS Best Newcomer Award
PODS 2026, 1–3 June 2026 (part of SIGMOD/PODS), Bengaluru, India

Near-Optimality for Single-Source Personalized PageRank, by Xinpeng Jiang, Haoyu Liu, and Siqiang Luo, of the College of Computing and Data Science at Nanyang Technological University, together with Prof Xiao, won the Best Newcomer Award at the ACM Symposium on Principles of Database Systems. The paper largely closes a long-standing gap between the best known upper and lower bounds on how efficiently personalized PageRank, a graph-ranking method used in web search and recommendation systems, can be computed for a single source node.