Theoretical and practical advances in preprocessing based secure computation

Agarwal, Amit

Permalink

https://hdl.handle.net/2142/130186 Copy

Description

Title

Theoretical and practical advances in preprocessing based secure computation

Author(s)

Agarwal, Amit

Issue Date

2025-07-16

Director of Research (if dissertation) or Advisor (if thesis)

Khurana, Dakshita

Doctoral Committee Chair(s)

Khurana, Dakshita

Committee Member(s)

• Miller, Andrew
• Gunter, Carl
• Beaver, Donald

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Secure Computation
• Zero-Knowledge Proofs
• Cryptography

Abstract

Secure computation — often called multiparty computation (MPC) — is a cornerstone of modern cryptography, enabling multiple parties to jointly compute functions over their private inputs without revealing those inputs. Over the past two decades, the preprocessing model of MPC has emerged as a powerful paradigm for improving practical efficiency. In this approach, the protocol is split into two distinct phases: 1.) Offline (input-independent) phase: Parties perform the bulk of the cryptographic work ahead of time to generate “correlated randomness.” 2.) Online (input-dependent) phase: Parties consume that precomputed “correlated randomness” to execute the actual secure computation task in an efficient way. By shifting intensive computations to the offline phase, the online phase can run with minimal latency, significantly reducing the response time of the protocol. This dissertation tackles the two core challenges of preprocessing-based MPC: 1.) Choosing the right correlated randomness: We study two important applications — secure sorting and secure logistic regression — and identify specialized forms of correlated randomness that yield communication-efficient online protocols for each task. 2.) Generating and storing correlated randomness efficiently: We introduce new techniques for producing two key types of correlations — unit-vector correlations and doubly-authenticated bits — by harnessing pseudorandom generators with enhanced properties. We further demonstrate how these correlations accelerate secure computation and zero-knowledge proofs respectively. Together, these contributions advance the state of the art in MPC by both broadening the range of efficiently. solvable tasks and streamlining the resources required to prepare for them.

Graduation Semester

2025-08

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/130186

Copyright and License Information

Copyright 2025 Amit Agarwal

Owning Collections