Welcome to the resource topic for 2018/471
Title:
Efficient Range ORAM with \mathbb{O}(\log^{2}{N}) Locality
Authors: Anrin Chakraborti, Adam J. Aviv, Seung Geol Choi, Travis Mayberry, Daniel S. Roche, Radu Sion
Abstract:Oblivious RAM protocols (ORAMs) allow a client to access data from an untrusted storage device without revealing to that device any information about their access pattern. Typically this is accomplished through random shuffling of the data such that the storage device cannot determine where individual blocks are located, resulting in a highly randomized access pattern. Storage devices however, are typically optimized for \emph{sequential} access. A large number of random disk seeks during standard ORAM operation induce a substantial overhead. In this paper, we introduce rORAM, an ORAM specifically suited for accessing ranges of \emph{sequentially logical blocks} while \emph{minimizing the number of random physical disk seeks}. rORAM obtains significantly better asymptotic efficiency than prior designs (Asharov et al., ePrint 2017, Demertzis et al., CRYPTO 2018) reducing {\em both} the number of seeks and communication complexity by a multiplicative factor of \mathbb{O}(\log N). An rORAM prototype is 30-50x times faster than Path ORAM for similar range-query workloads on local HDDs, 30x faster for local SSDs, and 10x faster for network block devices. rORAM’s novel disk layout can also speed up standard ORAM constructions, e.g., resulting in a 2x faster Path ORAM variant. Importantly, experiments demonstrate suitability for real world applications – rORAM is up to 5x faster running a file server and up to 11x faster running a range-query intensive video server workloads compared to standard Path ORAM.
ePrint: https://eprint.iacr.org/2018/471
See all topics related to this paper.
Feel free to post resources that are related to this paper below.
Example resources include: implementations, explanation materials, talks, slides, links to previous discussions on other websites.
For more information, see the rules for Resource Topics .