INTRODUCTION
CSPs provide desirable service properties, such as scalability, elasticity, fault
tolerance, and pay per use. Thus, cloud computing has become a promising service
paradigm to support IoT applications and IoT system deployment. To ensure data
privacy, existing research proposes to outsource only encrypted data to CSPs.
However, the same or different users could save duplicated data under different
encryption schemes at the cloud. Although cloud storage space is huge, this kind of
duplication wastes networking resources, consumes excess power, and complicates
data management. Thus, saving storage is becoming a crucial task for CSPs.
Deduplication can achieve high space and cost savings, reducing up to 90 to 95
percent of storage needs for backup applications ([Link] and up to
68 percent in standard file systems.1 Obviously, the savings, which can be passed
back directly or indirectly to cloud users, are significant to the economics of cloud
business. At the same time, data owners want CSPs to protect their personal data
from unauthorized access. CSPs should therefore perform access control based on
the data owner’s expectations. In addition, data owners want to control not only
data access but also its storage and usage. From a flexibility viewpoint, data
deduplication should cooperate with data access control mechanisms. That is, the
same data, although in an encrypted form, is only saved once at the cloud but can
be accessed by different users based on the data owners’ policies. However,
current industrial deduplication solutions can’t handle encrypted data. Existing
solutions for deduplication are vulnerable to brute-force attacks2 and can’t flexibly
support data access control and revocation (see the “Related Work in Data
Deduplication” sidebar for a discussion of some other work in this area).3 Few
existing schemes for cloud data access control support data deduplication
simultaneously, 4 and few can ensure flexibility and security with sound
performance for cloud data deduplication that data owners control directly.5–7 We
propose a scheme based on attribute-based encryption (ABE) to deduplicate
encrypted data stored in the cloud and support secure data access ontrol at the same
time. Analysis and implementation demonstrate that our scheme is secure,
effective, and efficient.
System Design and Algorithms
Our scheme works in a system containing three types of entities:
• a CSP that offers a storage service and performs honestly on data storage and
management to gain commercial profit but can’t be fully trusted since it’s curious
about the contents of stored data;
• a data owner that stores its data at the CSP (we assume there’s only one data
owner for one data M); and
• data holders (ui, i = 1, . . . , n) that are eligible data users and could save the same
data as the data owner at the CSP.
We assume that the system uses a data owner or holder device (such as a smart
mobile device) to collect and preprocess data collected by IoT devices if they
aren’t capable of cryptographic operations or networking. Additional assumptions
include the data’s hash ncode M(H(M)) being applied as its indicator, which is
used to check the duplication of data during data storage. We assume the data
holder signs the right hash code honestly for ownership verification at the CSP.
This hash code is protected and can’t be obtained by attackers. We further assume
that the data owner has the highest priority for data storage management. A data
holder should provide valid proof of ownership to request special treatment. The
CSP, data owners, and data holders communicate with each other through a secure
channel. System Setup and Required Keys In our scheme, we manage the keys of
different authorized data users for deduplication by treating their identities (IDs) as
valid attributes for accessing encrypted data stored at the CSP. Our scheme is
applicable to scenarios in which the data owner wants to control data storage and
access and track its own data location and usage status. During system setup, every
data owner or holder, CSP user u, maintains a public key PKu, which other users
employ to generate personalized secret attribute keys, and a secret key SKu, which
is used in the decryption operation related to PKu. The IEEE Cloud
Computing,Volume:3,Issue:2,Mar.-Apr.2016. data owner or holder uses the SKu to
issue secret attribute keys to other users and to generate its own public key PKIDu
of identity attribute ID. CSP user u generates PKIDu to encrypt a symmetric key
DEKu, randomly selected for encrypting the data of u, aiming to control data
access and deduplication. The corresponding secret attribute keys for decrypting
the cipher key encrypted by PKIDu are personalized for eligible data holders and
issued by data owner u. To prevent collusion, every holder gets a different secret
attribute key that only it can use. A secret attribute key of the attribute ID, issued
for an eligible data holder u′ by u is denoted as SKID(u,u′). Meanwhile, user u also
generates a key pair pku and sku for Public-Key Cryptosystem (PKC), for example,
signature generation and verification. The keys (PKu, SKu), (pku, sku) are bound to
the unique identity of u, which can be a unique anonymous identifier. This binding
is crucial for user identity verification. At system setup, PKu and pku are certified
by an authorized third party as Cert(PKu), Cert(pku) which the CSP and any CSP
users can verify. Algorithms Our scheme consists of several fundamental
algorithms. We can adopt either cyphertext policy ABE (CP-ABE) or key policy
ABE (KP-ABE) to implement related algorithms.