Research

My current research looks at how to use tools from theoretical CS, particularly cryptography, to support accountability in data sharing systems, including statistical data analysis and machine learning. In undergrad, I did research in computational complexity theory, specifically within proof complexity and meta-complexity. Underlying my research is an abiding interest in TCS methodology, namely: what makes a mathematical abstraction "good," and what are useful methods for designing good abstractions?

Bell, Z. R., A. Thudi, O. Franzese-McLaughlin, N. Papernot, S. Goldwasser. “Efficient Public Verification of Private Machine Learning via Regularization” In submission (2025): https://arxiv.org/abs/2512.04008. 

The public currently lacks an efficient way to verify that models trained on their data satisfy differential privacy (DP) guarantees, as existing methods require computation that scales with that required to train. We design the first DP algorithm with near optimal privacy-utility trade-offs but whose DP verification is cheaper than training. This leads to significantly reduced DP verification costs for stochastic convex optimization on large datasets—e.g. from 100 to only 3 hours on MNIST—which any member of the public can check.

Bell, Z. R. “Going Meta on the Minimum Circuit Size Problem: How Hard Is It to Show How Hard Showing Hardness Is?” HMC Senior Theses (2021), 250: https://scholarship.claremont.edu/hmc_theses/250.

Bell, Z. R., J. Camero, K. Cho, T. Hyde, C. Lu, R. Miller, B. Thompson, E. Zhu. “Density of Periodic Points for Lattès Maps over Finite Fields.” The Journal of Number Theory (2022), Volume 238 p. 951-966: https://www.sciencedirect.com/science/article/abs/pii/S0022314X2100367X.