Content area

Abstract

Privacy-preserving machine learning, especially through homomorphic encryption (HE), is essential to facilitate secure data processing in AI-driven enterprise systems. Nevertheless, HE's computational burden significantly restricts its practical implementation, particularly in critical domains like healthcare and edge devices, where data security and processing efficiency are essential. This study investigates a caching solution to mitigate this bottleneck, focusing on caching radix trees as a means to enhance the computational efficiency of HE-based processes.

The research introduces and evaluates a caching approach that uses currently available homomorphic encryption libraries. The aim is to reduce the processing time of homomorphic encryption operations with the goal of making privacy-preserving machine learning applications more scalable. This optimization method deliberately caches ciphertexts, facilitating a transition from quadratic to almost linear computational complexity. Key performance indicators are examined, including the inference time and correctness of the computation across various HE schemes and compute configurations.

The research findings indicate that radix tree caching can substantially improve HE performance, making it suitable for wider implementation in enterprise settings. This contribution tackles the efficiency difficulties of HE while creating a framework for privacy-preserving approaches that facilitate real-time, data-intensive applications without jeopardizing security or accuracy. This practice offers a pragmatic solution to a significant obstacle in privacy-preserving machine learning, enhancing both theoretical and applied understanding in secure AI.

Details

1010268
Business indexing term
Title
Advancing Adoption of Privacy Preserving Machine Learning in AI-Enabled Enterprises: Efficient Homomorphic Encryption Through Radix Caching
Author
Number of pages
99
Publication year
2025
Degree date
2025
School code
0075
Source
DAI-B 87/2(E), Dissertation Abstracts International
ISBN
9798290963150
Committee member
Etemadi, Amir; Akinfaderin, Adewale
University/institution
The George Washington University
Department
Computer Science
University location
United States -- District of Columbia
Degree
D.Engr.
Source type
Dissertation or Thesis
Language
English
Document type
Dissertation/Thesis
Dissertation/thesis number
32170898
ProQuest document ID
3240385421
Document URL
https://www.proquest.com/dissertations-theses/advancing-adoption-privacy-preserving-machine/docview/3240385421/se-2?accountid=208611
Copyright
Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.
Database
ProQuest One Academic