Joseph Near

Joseph Near
Nationality	American
Occupation	Computer scientist, academic
Employer	University of Vermont
Known for	Differential privacy, programming languages for privacy and security
Alma mater	University of Washington

Joseph Near is a computer scientist and associate professor in the Department of Computer Science at the University of Vermont (UVM). His research centers on the intersection of programming languages, formal verification, and data privacy, with particular emphasis on differential privacy and secure multiparty computation. Over the course of his career he has contributed to the development of programming-language tools that make it easier for practitioners to build systems that provably protect individual privacy while still allowing useful statistical analysis of sensitive data.

Education

Near received a Bachelor of Science in computer science from Indiana University, and a Master of Science and a Ph.D. in computer science from the Massachusetts Institute of Technology.

Career

Near joined the faculty of the University of Vermont, where he holds a position in the Department of Computer Science within the College of Engineering and Mathematical Sciences. At UVM, he leads a research group that develops programming-language techniques, verification tools, and usable systems for enforcing differential privacy and secure computation guarantees in real-world software.

In April 2023, Near received a Faculty Early Career Development (CAREER) award from the National Science Foundation, one of the agency's most prominent honors for junior faculty. The award, titled "Distributed Differential Privacy via Secure Multiparty Computation," provided $433,738 in funding to the University of Vermont and State Agricultural College to support research combining differential privacy with secure multiparty computation (MPC) techniques. The project aims to design systems that allow multiple parties to jointly compute statistics over their combined data without a trusted central curator, addressing a long-standing gap between the theory of differential privacy, which typically assumes a trusted aggregator, and real-world deployments in which no single party can be fully trusted with raw data.

Near's professional activity spans collaborations with researchers in programming languages, cryptography, and applied security. He has co-authored work appearing in venues including the ACM Transactions on Programming Languages and Systems, the European Symposium on Programming, the North American Chapter of the Association for Computational Linguistics, and the USENIX Symposium on Usable Privacy and Security (SOUPS).

Research

Near's research program addresses the formal and practical challenges of building software systems that protect individual privacy. A central theme of his work is the use of programming-language techniques, including type systems, static analysis, and formal verification, to make it easier to construct differentially private algorithms correctly and to prove that such algorithms satisfy their intended privacy guarantees.

One strand of this research concerns the design of type systems and language constructs that track the "privacy cost" of computations as a program executes, allowing a compiler or type checker to certify that a program satisfies a specified differential privacy budget. His paper "Contextual Linear Types for Differential Privacy," published in ACM Transactions on Programming Languages and Systems in 2023, presents a type system that uses linear types to reason about the sensitivity of computations in a context-sensitive manner, and has been cited fourteen times according to Semantic Scholar.

A second strand of Near's research addresses the usability of differential privacy tools for practitioners who are not experts in privacy theory. His 2023 paper "Evaluating the Usability of Differential Privacy Tools with Data Practitioners," presented at the Symposium on Usable Privacy and Security (SOUPS) co-located with the USENIX Security Symposium, reports on empirical studies of how data analysts interact with differential privacy software and identifies usability barriers to adoption. That paper has been cited twenty-eight times, making it among his most cited works.

A third strand concerns secure multiparty computation (MPC) and its combination with differential privacy, an area supported by his NSF CAREER award. Related publications include "Language-Based Security for Low-Level MPC," presented at the ACM SIGPLAN International Conference on Principles and Practice of Declarative Programming in 2024, and "SMT-Boosted Security Types for Low-Level MPC," presented at the European Symposium on Programming in 2025. These works apply programming-language security techniques, including type systems and SMT-solver-assisted verification, to low-level implementations of secure multiparty computation protocols in order to detect information leaks and enforce security properties at compile time.

Near has also examined the interaction of differential privacy with machine learning, including how model initialization and training procedures affect the accuracy-privacy tradeoff in differentially private learning. His paper "Differentially Private Learning Needs Better Model Initialization and Self-Distillation," presented at the North American Chapter of the Association for Computational Linguistics (NAACL) in 2024, proposes techniques to improve the performance of differentially private machine learning models. More recent work, including "Differentially Private Multimodal In-Context Learning," posted to arXiv in advance of a 2026 publication, extends privacy-preserving techniques to multimodal machine learning systems and in-context learning settings used by large language models.

Near's contributions also extend into distributed systems for query answering under differential privacy, exemplified by his work on "Distributed HDMM: Scalable, Distributed, Accurate, and Differentially Private Query Workloads without a Trusted Curator," which builds on the High-Dimensional Matrix Mechanism (HDMM) to enable accurate differentially private answers to workloads of statistical queries in distributed settings.

In addition to these core areas, Near has contributed to survey and tutorial literature intended to make differential privacy accessible to a broader technical audience, including "Programming Differential Privacy," published in Foundations and Trends in Programming Languages in 2025, which synthesizes programming-language approaches to differential privacy for researchers and practitioners entering the field.

According to Semantic Scholar, Near has authored or co-authored fifty-two papers that have collectively been cited 1,390 times, yielding an h-index of 20.

Recognition

His scholarly output has also achieved measurable citation impact within the programming languages and privacy research communities, reflected in an h-index of 20 across his publication record as tracked by Semantic Scholar. His 2023 SOUPS paper on the usability of differential privacy tools and his ACM Transactions on Programming Languages and Systems paper on contextual linear types for differential privacy are, respectively, his two most cited works, with twenty-eight and fourteen citations.

Publications

Selected publications by Joseph Near include:

• Near, J. (2026). "Differentially Private Multimodal In-Context Learning." arXiv.org.
• Near, J. (2026). "Qoreo: Choreographic Programming for Quantum Distributed Systems."