adfa, p. 1, 2013.
© Springer-Verlag Berlin Heidelberg 2013
A Specific Encryption Solution for Data Warehouses
Ricardo Jorge Santos
1
, Deolinda Rasteiro
2
, Jorge Bernardino
3
and Marco Vieira
4
1, 4
CISUC – FCTUC – University of Coimbra – 3030-290 Coimbra – Portugal
2
DFM – ISEC – Polytechnic Institute of Coimbra – 3030-190 Coimbra – Portugal
3
CISUC – ISEC – Polytechnic Institute of Coimbra – 3030-190 Coimbra – Portugal
lionsoftware.ricardo@gmail.com, dml@isec.pt, jorge@isec.pt, mvieira@dei.uc.pt
Abstract. Protecting Data Warehouses (DWs) is critical, because they store the
secrets of the business. Although published research and best practice guides
state encryption is the best way to assure the confidentiality of sensitive data
and maintain high performance, this adds overheads that jeopardize their feasi-
bility in DWs. In this paper, we propose a Specific Encryption Solution tailored
for DWs (SES-DW), using a numerical cipher with variable mixes of eXclusive
Or (XOR) and modulo (division remainder) operators. Data storage overhead is
avoided by preserving each encrypted column’s datatype, while transparent
SQL rewriting is used to avoid I/O and network bandwidth bottlenecks by dis-
carding data roundtrips for encryption and decryption purposes. The experimen-
tal evaluation using the TPC-H benchmark and a real-world sales DW with
Oracle 11g and Microsoft SQL Server 2008 shows that SES-DW achieves bet-
ter response time in both inserting and querying, than standard and state-of-the-
art encryption algorithms such as AES, 3DES, OPES and Salsa20, while pro-
viding considerable security strength.
Keywords: Encryption, Confidentiality, Security, Data Warehousing.
1 Introduction
Data Warehouses (DWs) store extremely sensitive information used for producing
business knowledge and aiding decision support. Unauthorized disclosure is therefore,
a critical security issue. To avoid this, encryption is widely used. However, although
most encryption solutions provide high security strength, they also introduce very
high performance overheads, as shown in [16]. Since decision support queries usually
access huge amounts of data (ranging from few MegaBytes to many TeraBytes), re-
sulting in substantial response time (usually from minutes to hours) [12], the overhead
introduced by using encryption may be unfeasible for DW environments if they are
too slow to be considered acceptable in practice [13]. Thus, encryption solutions built
for DWs must balance security and performance tradeoff requirements, i.e., they must
ensure strong security while keeping database performance acceptable [13, 16].
As the number and complexity of “data-mix” encryption rounds increase, their se-
curity strength often improves while performance degrades, and vice-versa. Balancing
performance with security in real-world DW scenarios is a complex issue which de-
pends on the requirements and context of each particular environment. Most current
encryption algorithms are not suitable for DWs, because they have been designed as a
general-purpose “one fits all” security solution. This introduces a need for specific
solutions for DWs capable of producing better security-performance tradeoffs.