Paper 2023/578 Original: eprint.iacr.org/2023/578 ↗

DORAM revisited: Maliciously secure RAM-MPC with logarithmic overhead

Brett Falk
Daniel Noble
Rafail Ostrovsky
Matan Shtepel
Jacob Zhang
Abstract

Distributed Oblivious Random Access Memory (DORAM) is a secure multiparty protocol that allows a group of participants holding a secret-shared array to read and write to secret-shared locations within the array. The efficiency of a DORAM protocol is measured by the amount of communication and computation required per read/write query into the array. DORAM protocols are a necessary ingredient for executing Secure Multiparty Computation (MPC) in the RAM model. Although DORAM has been widely studied, all existing DORAM protocols have focused on the setting where the DORAM servers are semi-honest. Generic techniques for upgrading a semi-honest DORAM protocol to the malicious model typically increase the asymptotic communication complexity of the DORAM scheme. In this work, we present a 3-party DORAM protocol which requires $O((\kappa + D)\log N)$ communication and computation per query, for a database of size $N$ with $D$-bit values, where $\kappa$ is the security parameter. Our hidden constants in a big-O nation are small. We show that our protocol is UC-secure in the presence of a malicious, static adversary. This matches the communication and computation complexity of the best semi-honest DORAM protocol, and is the first malicious DORAM protocol with this complexity.

Metadata
Available format(s)
PDF
Category
Cryptographic protocols
Publication info
Preprint.
History
2023-12-23: last updated
2023-04-24: published
Short URL
https://ia.cr/2023/578
License
https://creativecommons.org/licenses/by/4.0/

BibTeX

@misc{cryptoeprint:2023/578,
      author = {Brett Falk and Daniel Noble and Rafail Ostrovsky and Matan Shtepel and Jacob Zhang},
      title = {DORAM revisited: Maliciously secure RAM-MPC with logarithmic overhead},
      howpublished = {Cryptology {ePrint} Archive, Paper 2023/578},
      year = {2023},
      url = {https://eprint.iacr.org/2023/578}
}
Note: In order to protect the privacy of readers, this archive does not use cookies or embedded third party content.