TY - GEN
T1 - Model Parameter Estimation and Inference on Encrypted Domain
T2 - 18th World International Conference on Information Security and Application, WISA 2017
AU - Lee, Saetbyeol
AU - Yoon, Jiwon
N1 - Funding Information:
Acknowledgments. This research was supported by the MSIP (Ministry of Science, ICT & Future Planning), Korea, under the IITP (Institute for Information & communications Technology Promotion) support program (2017-0-00545).
Publisher Copyright:
© 2018, Springer International Publishing AG, part of Springer Nature.
PY - 2018
Y1 - 2018
N2 - 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.
AB - 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.
KW - Cloud security
KW - Homomorphic encryption
KW - Image processing
KW - Leveled fully homomorphic encryption
KW - Statistical analysis
UR - http://www.scopus.com/inward/record.url?scp=85049487943&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-93563-8_10
DO - 10.1007/978-3-319-93563-8_10
M3 - Conference contribution
AN - SCOPUS:85049487943
SN - 9783319935621
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 115
EP - 126
BT - Information Security Applications - 18th International Conference, WISA 2017, Revised Selected Papers
A2 - Kang, Brent ByungHoon
A2 - Kim, Taesoo
PB - Springer Verlag
Y2 - 24 August 2017 through 26 August 2017
ER -