Embedding cryptographic features in compressive sensing
Yushu Zhang
a,b,c,
n
, Jiantao Zhou
b
, Fei Chen
c
, Leo Yu Zhang
b,c
, Kwok-Wo Wong
d
, Xing He
a
,
Di Xiao
e
a
Chongqing Key Laboratory of Nonlinear Circuits and Intelligent Information Processing, College of Electronic and Information Engineering, Southwest
University, Chongqing 400715, China.
b
Department of Computer and Information Science, Faculty of Science and Technology, University of Macau, Macau
c
College of Computer Science and Engineering, Shenzhen University, Shenzhen 518060, China
d
Department of Electronic Engineering, City University of Hong Kong, Kowloon, Hong Kong
e
College of Computer Science, Chongqing University, Chongqing 400044, China
article info
Article history:
Received 23 August 2015
Received in revised form
19 January 2016
Accepted 5 April 2016
Communicated by W.K. Wong
Available online 13 May 2016
Keywords:
Secure compressive sensing
Symmetric-key cipher
Parallel compressive sensing
Random permutation
abstract
Compressive sensing (CS) has been widely studied and applied in many fields. Recently, the way to
perform secure compressive sensing (SCS) has become a topic of growing interest. The existing works on
SCS usually take the sensing matrix as a key and can only be considered as preliminary explorations on
SCS. In this paper, we firstly propose some possible encryption models for CS. It is believed that these
models will provide a new point of view and stimulate further research in both CS and cryptography.
Then, we demonstrate that random permutation is an acceptable permutation with overwhelming
probability, which can effectively relax the Restricted Isometry Constant for parallel compressive sensing.
Moreover, random permutation is utilized to design a secure parallel compressive sensing scheme.
Security analysis indicates that the proposed scheme can achieve the asymptotic spherical secrecy.
Meanwhile, the realization of chaos is used to validate the feasibility of one of the proposed encryption
models for CS. Lastly, results verify that the embedding random permutation based encryption enhances
the compression performance and the scheme possesses high transmission robustness against additive
white Gaussian noise and cropping attack.
& 2016 Elsevier B.V. All rights reserved.
1. Introduction
Making use of the sparseness of natural signals, compressi ve sen-
sing (CS) [6,15,9,8] unifies sampling and compression to reduce the
data acquisition and computational load of the encoder , at the cost of a
higher computational complexity at the decoder. If the CS framework
can integrate with certain cryptogr aphic features for simultaneous
sampling, compression and encryption, its application areas can be
extended to, for example, limited-resource sensor and video surveil-
lance. It has been suggested in [8] that CS framework leads to an
encryption scheme, where the sensing matrix can be consider ed as an
encryption key . In recent years, there exist some pioneer works on
secure compressi ve sensing (SCS) [25,24,20,12,31,1–5]. Rachlin and
Baron [25] foundthatCScannotachieveperfectsecrecybutcan
guarantee computational secrecy . The definition of perfect secrecy [29]
requ ires that the occurrence probability of a message conditioned on
the cryptogram is equal to the aprioriprobability of the message,
PðX ¼xjY ¼yÞ¼PðX ¼ xÞ. Alternatively , the mutual information
satisfies IðX; YÞ¼0. In contrast to perfect secrecy , computational
secrecy relies on the difficulty in solving a hard computational pro-
blem (e.g. NP-hard) at the computation resources available to the
adversary . Orsdemir et al. [24] investigated the security and robustne ss
of employing a secret sensing matrix. They evaluated the security
against brute force and structured attacks. The analyses indicate that
the computational complexity of these attacks renders them infeasible
in practice. In addition, this SCS method w as found to have fair
robus tness agains t additiv e noise, maki ng it a promisin g encryp tion
techniqu e for multimedia applications. Hossein et al. [20] also
addressed the perfect secr ecy problem for the scenario that the
measurement matrix as a key is known to both the sender and the
receiver. Similar results have been obtained, as reported in [25].Itis
shown that the Shannon perfect secrecy is, in general, not achievable
by such a SCS method while a weaker sense of perfect secrecy may be
achieved under certain conditions. Agra wal and Vishwanath[1 ]
employ ed the CS frame wor k to establish secure phy sical lay er com-
munication ov er a Wyner wiretap channel. They showed that CS can
exploit channel asymmetry so that a message that is encoded as a
sparse vector is decodable with high probability at the receiver while it
is impossible to decode with high probability by the eavesdr opper .
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/neucom
Neurocomputing
http://dx.doi.org/10.1016/j.neucom.2016.04.053
0925-2312/& 2016 Elsevier B.V. All rights reserved.
n
Corresponding author at: Chongqing Key Laboratory of Nonlinear Circuits and
Intelligent Information Processing, College of Electronic and Information Engi-
neering, Southwest University, Chongqing 400715, China.
E-mail address: yushuboshi@163.com (Y. Zhang).
Neurocomputing 205 (2016) 472–480