Cryptanalysis of an image encryption scheme based on a new total shuffling algorithm

Chaotic systems have been broadly exploited through the last two decades to build encryption methods. Recently, two new image encryption schemes have been proposed, where the encryption process involves a permutation operation and an XOR-like transfo…

Authors: ** - 논문 원문에 저자 정보가 명시되어 있지 않음. (※ 필요 시 원 논문(

Cryptanalysis of an image encryption scheme based on a new total   shuffling algorithm
Cryptanalysis of an image encryptio n sc heme based on a new total sh uffling algorithm Da vid Arroy o a , ∗ , Chengqi ng Li b , Sh ujun Li c , Gonzalo Alv arez a and W olfga ng A. Halang c a Instituto de F ´ ısic a Aplic ada, Consejo Sup erior de Investigaciones Cient ´ ıfic as, Serr ano 144, 28006 Madrid, Sp ain b Dep artment of Ele ctr onic Engine ering, City University of H ong K ong, 83 T at Che e Avenue, Kow lo on T ong, Hong Kong SA R, China c F ernUniversit¨ at in Hagen, Chair of Computer E ngine ering, Universit¨ atsstr aße 27, 58084 Hagen, Germany Abstract Chaotic systems ha v e b een broadly exploited through the last t wo decades to build encryption metho ds. Recentl y , t wo new image en cr y p tion sc hemes hav e been pro- p osed, where the encryption pro cess in v olv es a p erm utation oper ation and an X OR- lik e transform ation of the sh uffled pixels, w h ic h are con trolled by three c haotic sys- tems. This pap er discusses some d efects of th e sc hemes and h o w to br eak them with a c hosen-plain text attac k. Key wor ds: Chaotic en cryption, Loren z system, Chen’s sy s tem, h yp er-chao s, logistic map, c hosen-plain text attac k, p erm utation-only encryption algorithms, cryptanalysis P ACS: 05.45.Ac, 47.20.Ky . 1 In tro duction When w e t hink ab out exc hanging information w e are ve ry inte rested in find- ing a w a y to make it fast and secure. Mo dern telecomm unications technolo- gies allo w to send and receiv e files, images, and data in a relativ ely short time dep ending on the bandwidth a v a ilable. Now ada ys, the use of traditional ∗ Corresp ond ing author: Da vid Arro y o (david.arro y o@iec.c sic.es). Preprint submitted to Ph ysics Letter A 2 No v em b er 2018 symmetric and asymmetric cryptography is the wa y to secure the info rma- tion exc hange [1 , 2]. How ev er, applications in v olving digita l images and videos demand other encryption sc hemes . Indeed, the b ulky size and the large re- dundancy of uncompres sed videos/images make it nec essary to lo ok for new metho ds to deal with those features in order to facilitate the in tegration o f the encryption in the whole pro cessing pro cedure. F or recen t surv eys on image and video encryption, please refer to [3–6]. The main features of c haotic systems (sensitivit y to initial conditions, ergo d- icit y , mixing prop ert y , simple analytic desc ription and high complex b eha vior) mak e them v ery in teresting to design new cryptosystems. Image encryption is an area where c haos has b een bro adly exploited. In fact, c haotic systems ha v e b een used to mask plain-imag es through XOR-lik e substitution op era- tions [7], spatial p ermutation [8] or the com bination of b ot h tec hniques [9]. This pap er is fo cused o n t w o image encryption sche mes prop osed in [1 0 , 11]. In b oth pap ers the image encryption is based on a secret p erm utation deriv ed from the logistic map, and a masking of the gray-scale v a lues of the sh uffled pixels with a k eystream generated from one or tw o c haotic systems. The only difference b etw een the t w o encryption sc hemes is that in [10] tw o c haotic sys- tems (Lorenz and Chen’s systems) are used to generate the key stream, while in [11] only one h yp er-c haotic system is used. Because suc h a difference is indep enden t of the securit y , w e only fo cus on the cryptanalysis of the sc heme prop osed in [10]. The rest of this pap er is organized as follo ws. The sc heme under study is described briefly in the next section. In Sec. 3 some imp ortan t problems of the cryptosystem are remark ed. Then, a ch osen-plainte xt at tac k is describ ed in Sec. 4 along with some experimental results. In the last section the conclusion is g iven. 2 The encryption sc heme Assuming that the size o f t he plain-imag e I is M × N and the cipher-image is I ′ , the encryption sc heme proposed in [1 0] can be describ ed b y the f o llo wing t w o pro cedures. Please note that w e use different notations from the original ones in [10] to get a simpler a nd clearer description. • Shuffling pr o c e dur e In this pro cedure, the plain- ima g e I is p erm uted to form an in termedi- ate image I ∗ according to a tot a l sh uffling matrix P ∗ , whic h is deriv ed b y pseudo-randomly p erm uting the rows and columns of the original p osition matrix P = [( i, j )]. The pseudo-random r ow and column p erm utations are generated by iterating the logistic map x n +1 = 4 x n (1 − x n ) from a giv en 2 initial condition x 0 . • Masking p r o c e dur e In this pro cedure, the in termediate image I ∗ is further mask ed b y a k eystream { B ( i ) } M N i =1 as follo ws: ∀ i = 1 ∼ M N , I ′ ( i ) = I ∗ ( i ) ⊕ B ( i ) ⊕ I ′ ( i − 1), where I ( i ), I ′ ( i ) denote the i -th pixels of I ∗ and I ′ (coun ted from left to right and from top to b ottom), respective ly , and I ′ (0) = 128. The k eystream { B ( i ) } M N i =1 is generated b y it era t ing the Lorenz and Chen’s systems and doing some p ostpro cessing on all the 6 c hao t ic v aria bles (the first N 0 iterations of Lo renz system a nd the first M 0 iterations of Chen’s systems are discarded to enhance t he securit y). Because our cryptanalysis succeeds regardless of the k eystream’s generation p ro cess, w e ignore this part and readers are referred to Sec. 2.3 of [10] fo r details. In [1 0], it is claimed that the secret ke y includes the initia l v alues of t he Lo r enz and Chen’s systems and the n um b er of initial iterations N 0 , M 0 . It is quite strange wh y the initial condition of the logistic map is not claimed to b e part of the k ey , since t he image encryption sc heme is based on “a new to t al sh uffling algorithm” (a s can b e seen in the title of [1 0]). In this cryptanalysis pap er, w e assume that the initial condition of the log istic map is also part o f the key . W e b eliev e it is also the original in ten tion of the authors of [10]. In addition, note that b oth P ∗ and { B ( i ) } M N i =1 are indep enden t of the plaintext and ciphertext, so they can b e used as an equiv alent key . 3 Design weaknesse s In t his section, w e discuss some defects of the sc heme under study . 3.1 L ow sensi tivity to the change of plain-image It is we ll know n that the ciphe rtext o f a secure encryption sc heme should b e v ery sensitiv e to the c hange of plain text [1 2 , Rule 9]. Unfortunately , the encryption sc heme under study fails to satisfy this req uiremen t. Giv en t w o plain-images I 0 and I 1 with only one pixel differenc e at the position ( i, j ), the difference will b e p erm uted to a new p osition ( i ∗ , j ∗ ) according to the sh uffling matr ix P ∗ . Then, b ecause all plain-pixels b efore ( i ∗ , j ∗ ) are iden tical for the t w o plain-images, the ciphertexts will also b e ide n tical. This show s the low sensitivit y o f the image encryption sch eme to changes in the plain- image. Fig ure 1 giv es an ex ample of this problem. It can b e seen ho w the differen tial cipher-image is equal to zero for any pixel b efore ( i ∗ , j ∗ ) and equal to a constant v a lue after that p osition. 3 (a) (b) (c) Fig. 1. Illu s tration of the lo w sensitivit y to the c hange of the plain-image: (a) the first plain-image I 0 ; (b) the second plain-image I 1 (only the cen ter pixel is different from I 0 ); (c) the d ifferen tial cipher-image I ′ 0 ⊕ I ′ 1 . 3.2 R e duc e d Key sp ac e As claimed in [10], N 0 and M 0 are also part of the k ey . Ho w ev er, from an attac k er’s p o in t of view, he/she only needs to guess the chaotic states after the N 0 and M 0 c haotic iterations as the initial conditions o f the L o renz a nd Chen’s systems. In this w a y , N 0 and M 0 are remo v ed from the k ey and the k ey space is reduced. 4 3.3 Pr oblem with chaotic iter ations o f L or enz and Ch en ’s systems In [10], the a utho rs did not sa y an ything ab out the time step τ of iterating the Lorenz and Chen’s systems. Ho w ev er, the randomness of the key stream { B ( i ) } M N i =1 is tightly dep enden t on the v a lue of time step. As an extreme exam- ple, if τ = 10 − 20 , w e will get a k eystream of identical elemen ts (according to the algorithm describ ed in Sec. 2.3 o f [10]) . As a matter of fact, the v alue o f τ is de- p enden t o n the m ultiplication facto r 10 13 o ccurring in Step 4 of the encryption pro cess (see Sec. 2.3 of [10]): x i = mo d ((abs( x i ) − Flo or(abs( x i ))) × 1 0 13 , 2 56). 3.4 L ow encryption sp e e d Because the chaotic iterations of Lorenz and Chen’s sys tems in v o lv e compli- cated numerical differential functions, the encryption sp eed is expected to b e v ery slo w compared with other traditional ciphers. T o asses this fact, w e de- riv ed a mo dified enc ryption sc heme from the original one b y replacing the Lorenz and Chen’s systems with the logistic map, and then compared the encryption sp eeds of the tw o cryptosystems . Both cryptosys tems w ere im- plemen ted using MA TLAB on a PC with a 1.6GHz pro cessor and 512MB of RAM. F or images of size 2 56 × 256, the t ypical encryption time f or the original cryptosystem in [10] w as around 5.8 sec onds, while t he mo dified cryptosystem based on the logistic map required in av erage around 1.2 se conds to encrypt an image. The exp erimen ts hav e clearly sho wn that using con tin uous c haotic systems can dr a stically reduce the encryption speed. Since there are also no other obv ious merits in using contin uous c haotic systems rather than a simple discrete-time c haotic map, the use o f the Lorenz and Chen’s sy stems in the image encryption sc heme under study is unnecessary . Inste ad, these con tin u- ous c haotic sys tems can b e replaced b y a simpler disc rete-time chaotic map without compromising the security . 4 Chosen-plain text att ac k When a v aria tion of stream cipher is created, as in the case under study , obtaining the k eystream is totally equiv alen t to obtaining the k ey whenev er differen t plain-images are encrypted using the same k ey . In this section, w e presen t a c hosen-plain text atta c k whic h a llo ws to reco v er b oth the k eystream and the shufflin g mat rix. Let us c ho ose a plain-image I 1 suc h tha t ∀ i, j = 1 ∼ M N , I 1 ( i ) = I 1 ( j ) = a . In this case, the sh uffling part does not w ork, so w e ha v e I ∗ 1 = I 1 . Then, w e can 5 reco v er the k eystream as follows : ∀ i = 1 ∼ M N , B ( i ) = I 1 ( i ) ⊕ I ′ 1 ( i ) ⊕ I ′ 1 ( i − 1). After remo ving the masking part, we can try to reco v er the sh uffling matrix. According to the general cryptanalysis on p ermutation-only ciphers in [13], only ⌈ log 256 ( M N ) ⌉ c hosen plain-images are needed to recov er the sh uffling matrix P ∗ . In total w e need ⌈ log 256 ( M N ) ⌉ + 1 c hosen plain-images to p erform this chose n-plaintext attac k. With the aim of v erifying t he prop osed attack , sev eral exp erimen ts hav e b een done. One o f the examples is shown in Fig. 2, where the images are of size 256 × 256 and the secret k ey in v olv ed is sho wn in T able 1. As it was mentioned ab ov e, the sh uffling pro cess is brok en usin g log 256 ( M N ) = 2 c hosen plain-images, while the masking pro cedure cryptanalysis requires one c hosen plain-image. The three c hosen plain-images allow to decipher the cipher-image included in Fig. 2(a) and thus to get the corresp onding plain- image (Fig. 2(b)), ev en when the secret ke y is unkno wn. T able 1 Key v alue used in the exp erimen t. x 1 (0) x 2 (0) x 3 (0) x 4 (0) x 5 (0) x 6 (0) N 0 M 0 x 0 0 . 3 − 0 . 4 1 . 2 10 . 2 − 3 . 5 4 . 4 3000 2000 0 . 4 (a) (b) Fig. 2. The resu lt of the c hosen-plaint ext attac k: (a) a cipher-image encrypted with the ke y as sh o wn in T able 1; (b) the decryp ted plain-image using the equiv alen t k ey  P ∗ , { B ( i ) } M N i =1  obtained via the chosen-plain text attac k. 5 Conclusions The securit y o f the image enc ryption sc heme prop o sed in [10] has b een ana- lyzed in detail. The cryptanalytic results are also v alid f o r the other sche me prop osed in [11]. It has b een sho wn that the equiv alent secret key can b e re- co v ered in a chosen-plain text attac k with only ⌈ log 256 ( M N ) ⌉ + 1 chosen pla in- images. In addition, some other defects ha v e also b een distinguished in the sc heme under study . Among those defects, it is necessary to emphasize the one 6 concerning the encryption sp eed, since it informs ab o ut the non-con v enience of con tin uous-time chaotic systems for implemen ting fa st encryption pro cedures. The w eak securit y pro p erties frustrate the usage of the sc heme in practice. Ac kno wledgmen ts The w or k described is this le tter w as partially supp orted by Ministerio de Educaci´ on y Ciencia of Spain, Researc h Grant SEG2004-024 1 8. Shujun Li w as supp orted b y the Alexander v on Humboldt F oundation, G erman y . References [1] S. V. A.J. Menezes, P .C. v an Oorsc hot, Hand b o ok of Applied Cryptograph y , CR C Press, 1997. [2] B. Sc hneier, App lied Cr yptograph y , John Wiley , NY, USA, 1996. [3] S. Li, G. Chen, X. Zheng, Ch aos-based encryption for d igital images and videos, in: B. F ur ht, D. Kiro vski (Eds .), Multimedia Security Handb o ok, CR C Press, LLC, 200 4, Ch . 4, pp. 133–1 67, preprint a v alaible at h ttp://www.ho oklee.com/pub.h tml. [4] A. Uh l, A. P omm er , F r om Digital Ri ght s Manag emen t to Secured P ersonal Comm unication, Sp ringer, 2005. [5] B. F uhrt, E. Muharemagic, D. Socek, Image encryption a lgorithms, in: Multimedia Encryp tion and W atermarking, Spr inger, 2005, Ch. 5, pp. 79–120. [6] W. Zeng, H.Y u, C.-Y. Lin (Eds.), Multimedia Security T ec hnologies for Digital Righ ts Managemen t, Academic Press, 2006. [7] H.-C. Chen, J.-C. Y en, A new cryptograph y system and its VLSI r ealization, Journal of Systems Arc hitecture 49 (7-9) (2003 ) 355–36 7. [8] J.-C. Y en, J.-I. Guo, Efficien t hierarchica l c haotic image encryption algorithm and its VLSI realisation, IEE Pro ceedings - Vision, Image an d S ignal Pro cessing 147 (2) (2000) 167–175. [9] G. Chen, Y. Mao, C. K. Chui, A symm etric image encr y p tion sc heme based on 3d c haotic cat maps, Chaos, S olitons an d F ractals 21 (3) (2004 ) 749–76 1. [10] T . Gao, Z . Ch en, Image encryption based on a new total shuffling algorithm, Chaos, Solitons and F ractals 0 (2007) doi:10.101 6/j.c haos.2006.1 1.009. [11] T . Gao, Q. Gu, Z. C hen, A new image encryption alg orithm b ased on h ye r- c haos, Phys. Lett. A 0 (200 7) doi:10.10 16/j.ph ysleta.2007.07.040. 7 [12] G. Alv arez, S. Li, Some basic cryptographic requ iremen ts for c haos-based cryptosystems, Inte rnational Journ al of Bifurcatio n an d Ch aos 16 (8) (2006) 2129– 2151. [13] S . Li, C. Li, G. Chen, N. G. Bourbakis, K.-T. Lo, A general cryptanalysis of p ermutatio n-only multimedia encryption algorithms, Cr yptology ePrint Arc hiv e, Rep ort 2004/37 4, http ://eprint .iacr.or g/2004/374 (2004). 8

Original Paper

Loading high-quality paper...

Comments & Academic Discussion

Loading comments...

Leave a Comment