Privacy and Accuracy-Aware AI/ML Model Deduplication

Authors

• Hong Guan, Arizona State University
• Lei Yu, Rensselaer Polytechnic Institute
• Lixi Zhou, Arizona State University
• Li Xiong, Emory University
• Kanchan Chowdhury, Marquette UniversityFollow
• Lulu Xie, Arizona State University
• Xusheng Xiao, Arizona State University
• Jia Zou, Arizona State University

Document Type

Article

Publication Date

6-2025

Publisher

Academy of Management

Source Publication

Academy of Management Proceedings

Source ISSN

0065-0668

Original Item ID

DOI: 10.1145/3725340

Abstract

With the growing adoption of privacy-preserving machine learning algorithms, such as Differentially Private Stochastic Gradient Descent (DP-SGD), training or fine-tuning models on private datasets has become increasingly prevalent. This shift has led to the need for models offering varying privacy guarantees and utility levels to satisfy diverse user requirements. Managing numerous versions of large models introduces significant operational challenges, including increased inference latency, higher resource consumption, and elevated costs. Model deduplication is a technique widely used by many model serving and database systems to support high-performance and low-cost inference queries and model diagnosis queries. However, none of the existing model deduplication works has considered privacy, leading to unbounded aggregation of privacy costs for certain deduplicated models and inefficiencies when applied to deduplicate DP-trained models. We formalize the problem of deduplicating DP-trained models for the first time and propose a novel privacy- and accuracy-aware deduplication mechanism to address the problem. We developed a greedy strategy to select and assign base models to target models to minimize storage and privacy costs. When deduplicating a target model, we dynamically schedule accuracy validations and apply the Sparse Vector Technique to reduce the privacy costs associated with private validation data. Compared to baselines, our approach improved the compression ratio by up to 35× for individual models (including large language models and vision transformers). We also observed up to 43× inference speedup due to the reduction of I/O operations.

Comments

Academy of Management Proceedings, Vol. 3, No. 3 (June 2025). DOI.

Recommended Citation

Guan, Hong; Yu, Lei; Zhou, Lixi; Xiong, Li; Chowdhury, Kanchan; Xie, Lulu; Xiao, Xusheng; and Zou, Jia, "Privacy and Accuracy-Aware AI/ML Model Deduplication" (2025). Computer Science Faculty Research and Publications. 134.
https://epublications.marquette.edu/comp_fac/134