(Georgia Institute of Technology, 2014)
Tang, Yuzhe; Iyengar, Arun; Tan, Wei; Fong, Liana; Liu, Ling
The recent shift towards write-intensive workload on
big data (e.g., financial trading, social user-generated data
streams) has pushed the proliferation of the log-structured key-value stores, represented by Google’s BigTable, HBase
and Cassandra; these systems optimize write performance by
adopting a log-structured merge design. While providing key-based
access methods based on a Put/Get interface, these
key-value stores do not support value-based access methods,
which significantly limits their applicability in many web and Internet applications, such as real-time search for all tweets or blogs containing “government shutdown”. In this paper, we present HINDEX, a write-optimized indexing scheme on the log-structured key-value stores. To index intensively
updated big data in real time, the index maintenance is made
lightweight by a design tailored to the unique characteristic of the underlying log-structured key-value stores. Concretely, HINDEX performs append-only index updates, which avoids the reading of historic data versions, an expensive operation
in the log-structure store. To fix the potentially obsolete index entries, HINDEX proposes an offline index repair
process through tight coupling with the routine compactions. HINDEX’s system design is generic to the Put/Get interface;
we implemented a prototype of HINDEX based on HBase
without internal code modification. Our experiments show
that the HINDEX offers significant performance advantage for the write-intensive index maintenance.
(Georgia Institute of Technology, 2014)
Tang, Yuzhe; Liu, Ling; Iyengar, Arun
In information sharing networks, having a privacy preserving
index (or PPI) is critically important for providing
efficient search on access controlled content across distributed
providers while preserving privacy. An understudied
problem for PPI techniques is how to provide controllable
privacy preservation, given the innate difference of
privacy of the different content and providers. In this paper
we present a configurable privacy preserving index, coined
𝜖-PPI, which allows for quantitative privacy protection
levels on fine-grained data units. We devise a new common-identity
attack that breaks existing PPI’s and propose an
identity-mixing protocol against the attack in 𝜖-PPI. The
proposed 𝜖-PPI construction protocol is the first without
any trusted third party and/or trust relationship between
providers. We have implemented our 𝜖-PPI construction
protocol by using generic MPC techniques (secure multiparty
computation) and optimized the performance to a
practical level by minimizing the costly MPC computation
part.