Model Parameter Estimation and Inference on Encrypted Domain: Application to Noise Reduction in Encrypted Images

Saetbyeol Lee, Jiwon Yoon

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

One of the major issues in security is how to protect the privacy of multimedia big data on cloud systems. Homomorphic Encryption (HE) is increasingly regarded as a way to maintain user privacy on the untrusted cloud. However, HE is not widely used in machine learning and signal processing communities because the HE libraries are currently supporting only simple operations like integer addition and multiplication. It is known that division and other advanced operations cannot feasibly be designed and implemented in HE libraries. Therefore, we propose a novel approach to building a practical matrix inversion operation using approximation theory on HE. The approximated inversion operation is applied to reduce unwanted noise on encrypted images. Our research also suggests the efficient computation techniques for encrypted matrices. We conduct the experiment with real binary images using open source library of HE.

Original languageEnglish
Title of host publicationInformation Security Applications - 18th International Conference, WISA 2017, Revised Selected Papers
EditorsBrent ByungHoon Kang, Taesoo Kim
PublisherSpringer Verlag
Pages115-126
Number of pages12
ISBN (Print)9783319935621
DOIs
Publication statusPublished - 2018
Event18th World International Conference on Information Security and Application, WISA 2017 - Jeju Island, Korea, Republic of
Duration: 2017 Aug 242017 Aug 26

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10763 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other18th World International Conference on Information Security and Application, WISA 2017
Country/TerritoryKorea, Republic of
CityJeju Island
Period17/8/2417/8/26

Keywords

  • Cloud security
  • Homomorphic encryption
  • Image processing
  • Leveled fully homomorphic encryption
  • Statistical analysis

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint

Dive into the research topics of 'Model Parameter Estimation and Inference on Encrypted Domain: Application to Noise Reduction in Encrypted Images'. Together they form a unique fingerprint.

Cite this