Rate My Professor Ben Rubinstein

His research specializes in AI safety, covering security and privacy of AI systems through areas such as adversarial machine learning, differential privacy, and record linkage. Key publications include the book Adversarial Machine Learning (Cambridge University Press, 2019, co-authored with Anthony D. Joseph and Blaine Nelson); "Double Bubble, Toil and Trouble: Enhancing Certified Robustness through Transitivity" (NeurIPS 2022, oral presentation, with Andrew C. Cullen et al.); "Certified Adversarial Robustness via Randomized α-Smoothing for Regression Models" (NeurIPS 2024, with Aref Miri Rekavandi et al.); "Pain-Free Random Differential Privacy with Sensitivity Sampling" (ICML 2017, with Francesco Alda); and "Shifting, one-inclusion mistake bounds and tight multiclass expected risk bounds" (NIPS 2007, with Peter L. Bartlett and J. Hyam Rubinstein). Since 2013, he has attracted $11.75 million in competitive grants ($7.45 million as lead). Awards include the Australian Research Council DECRA Fellowship and the 2016 Young Tall Poppy Science Award (Victoria). His work has influenced policy, such as deanonymizing Myki and Medicare data leading to the Re-identification Offence Bill 2016; privacy audits for Australian Bureau of Statistics, National Australia Bank, and Transport for NSW; Firefox side-channel fixes; and scalable record linkage adopted by the U.S. Census Bureau. Rubinstein serves on the Australian Academy of Sciences National Committee for Information and Communication Sciences, Computing @ Go8, Kingston AI Group, and co-leads the CATCH MURI cybersecurity project.