Paper 2024/1534

More Efficient Lattice-based OLE from Circuit-private Linear HE with Polynomial Overhead

Leo de Castro, Massachusetts Institute of Technology
Duhyeong Kim, Intel (United States)
Miran Kim, Hanyang University
Keewoo Lee, University of California, Berkeley
Seonhong Min, Seoul National University
Yongsoo Song, Seoul National University
Abstract

We present a new and efficient method to obtain circuit privacy for lattice-based linearly homomorphic encryptions (LHE). In particular, our method does not involve noise-flooding with exponetially large errors or iterative bootstrapping. As a direct result, we obtain a semi-honest oblivious linear evaluation (OLE) protocol with the same efficiency, reducing the communication cost of the prior state of the art by 50%. Consequently, the amortized time of our protocol improves the prior work by 33% under 100Mbps network setting. Our semi-honest OLE is the first to achieve both concrete efficiency and asymptotic quasi-optimality. Together with an extension of the recent zero-knowledge proof of plaintext knowledge, our LHE yields actively-secure OLE with 2.7x reduced communication from the prior work. When applied to Overdrive (Eurocrypt '18), an MPC preprocessing protocol, our method provides 1.4x improvement in communication over the state of the art.

Metadata
Available format(s)
PDF
Category
Cryptographic protocols
Publication info
Preprint.
Keywords
Oblivious Linear EvaluationCircuit PrivacyMulti-Party ComputationHomomorphic Encryption
Contact author(s)
ldec @ mit edu
duhyeong kim @ intel com
miran @ hanyang ac kr
keewoo lee @ berkeley edu
minsh @ snu ac kr
y song @ snu ac kr
History
2024-10-04: approved
2024-10-01: received
See all versions
Short URL
https://ia.cr/2024/1534
License
Creative Commons Attribution
CC BY

BibTeX

@misc{cryptoeprint:2024/1534,
      author = {Leo de Castro and Duhyeong Kim and Miran Kim and Keewoo Lee and Seonhong Min and Yongsoo Song},
      title = {More Efficient Lattice-based {OLE} from Circuit-private Linear {HE} with Polynomial Overhead},
      howpublished = {Cryptology {ePrint} Archive, Paper 2024/1534},
      year = {2024},
      url = {https://eprint.iacr.org/2024/1534}
}
Note: In order to protect the privacy of readers, eprint.iacr.org does not use cookies or embedded third party content.