Indonesian J our nal of Electrical Engineering and Computer Science V ol. 21, No. 2, September 2021, pp. 1103 1112 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v21i2.pp1103-1112 r 1103 An image encryption algorithm with a no v el chaotic coupled mapped lattice and chaotic image scrambling technique Behrang Chaboki, Ali Shakiba Department of Computer Science, V ali-e-Asr Uni v ersity of Rafsanjan, Iran Article Inf o Article history: Recei v ed Jun 13, 2020 Re vised Aug 11, 2020 Accepted Aug 21, 2020 K eyw ords: Chaotic coupled lattice mapping Chaotic image encryption Chaotic image scrambling ABSTRA CT In this paper , we b uild a no v el chaotic coupled lattice mapping with positi v e L ya- puno v e xponent, and introduce a no v el chaotic image scrambling mechanism. Then, we propose a chaotic image encryption algorithm which uses the introduced chaotic coupled lattice mapping to apply permutation by iterati v ely applying the introduced chaotic image scrambling mechanism, and dif fusing the pix el v alues. W e use a sorting approach rather than quantizing the chaotic floating-point v alues to construct the dif fu- sion matrix. W e also study the security of the proposed algorithm concerning se v eral security measures including brute-force attack, dif ferential attack, k e y sensiti vity , and statistical attacks. Moreo v er , the proposed algorithm is rob ust ag ainst data loss and noise attacks. This is an open access article under the CC BY -SA license . Corresponding A uthor: Ali Shakiba Department of Computer Scince V ali-e-Asr Uni v ersity of Rafsanjan, Iran Email: ali.shakiba@vru.ac.ir 1. INTR ODUCTION In the current w orld, with the staggering speed of technology de v elopment in the era of digital com- puting and with the widespread application of the internet in our life, the application of digital color images become more and more ine vitable. F or instance, the application of digital images in medical imaging [6], social and personal li fe, military and other applications are almost clear to e v eryone. So, in some situations, such as medical images and military applications, the concept of pri v ac y and security become one of the important challenges. One strate gy for solving this problem is to use some encryption techniques so that the image becom es unreadable for an unauthorized person. F or approaching this problem, man y encryption techniques ha v e been proposed, such as compressi v e sensing [1], quantum theory [2], DN A coding [3], transform domains [4], ma- trix transforms [5] and chaotic systems [7–9]. In the last 20 years, the chaotic system encryption algorithms ha v e been the attention of researchers due to its immense inherent aspects such as initial criteria sensiti vity , un- predictability , and pseudorandomness. Ho we v er , image encryption techniques that are solely based on chaotic systems, ha v e been sho wn to be vulnerable ag ainst cipherte xt-only and chosen plainte xt attacks. Thus, the per - formance of a chaotic system and structure of encrypt ion algorithm plays an important role in the resistance of an encryption scheme to the common attacks. By combining an appropriate chaotic-based system, in terms of initial sensiti vity , unpredictability , and randomness with proper techniques of confusion and dif fusion, a rob ust encryption algorithm will be acquired. J ournal homepage: http://ijeecs.iaescor e .com Evaluation Warning : The document was created with Spire.PDF for Python.
1104 r ISSN: 2502-4752 The rest of the paper is or g anized as follo ws: In Section 2, we firs t construct a no v el chaotic coupled mapped lattice using tw o one-dimensional chaotic mappings and study its chaotic properties. Then, we gen- eralize an image scrambling method and mak e it chaotic. These tw o primiti v es are combined to propose an image encryption algorithm. The security of the proposed method is studied with respect to se v eral security measures in Section 3. Finally , a conclusion is dra wn in Section 4. 2. METHOD In this paper we use the logistic mapping in combination with the Chebyshe v mapping of the first type, which is abbre viat ed as Chebyshe v mapping. These mappings are combined with coupled mapped lattices, or CML for s hort, to construct a no v el chaotic mapping with greater chaotic performance. The logistic mapping is a polynomial defined as x i +1 = L ( ; x i ) = x i (1 x i ) where x i 2 [0 ; 1] for i = 0 ; 1 ; : : : and 2 (0 ; 4] . This mapping sho ws a chaotic beha vior for 2 (3 : 569 945 6 ; 4] and this beha vior impro v es as gets closer to 4 . The Chebyshe v mapping is another chaotic mapping which is defined mathematically as T n ( x ) = cos ( n arccos( x )) where x 2 [ 1 ; 1] . Moreo v er , there is an equi v alence recurrence relation used to define Chebyshe v mappings as T n ( x ) = 2 xT n ( x ) T n 1 ( x ) with initial conditions T 0 ( x ) = T 1 ( x ) = 1 . The CML originally uses the logistic mapping to generate sequences with the follo wing relation X ( i ) n +1 = (1 " ) f [ x n ( i )] + " 2 f f [ x n ( i + 1)] + f [ x n ( i 1)] g where " is the coupling parameter and f [ x ] denotes the logistic map [10]. In this paper , we propose a no v el chaotic map with e xcellent chaotic beha vior using the CML with the logistic and the Chebyshe v mappings. This no v el mapping is defined as y ( i ) n +1 = (1 " ) L ; T n +1 x ( i )  + " 2 L ; T n +1 x ( i +1)  + L ; T n +1 x ( i 1)  where the sequences x ( i ) j are computed as follo ws: - A random initial v alue s 0 2 ( 1 ; 1) is generated - A sequence of chaotic v alues s i = T ` ( s i 1 ) are generated for i = 1 ; : : : ; m + m 0 - W e use the last m v alues of this sequence as the sequence x 0 , i.e. x ( i ) 0 = s m 0 + i for i = 1 ; : : : ; m and x i - F or n 1 , we set x ( i ) n = y ( i ) n 1+max y ( i ) n , i.e. we normalize the sequence y n to f all within the range [ 1 ; 1] The proposed chaotic system has good chaotic properties, as it can be observ ed in Fig. 1 which illustrates its the L yaponuv e xponent. Note that the L yapuno v e xponent describes the rate of the con v er gence or di v er gence of trajectories and positi v e v alues of L yapuno v e xponent sho w chaotic beha vior [11]. Figure 1. Bifurcation diagram of the proposed mapping with starting point x 0 = 1 10 5 and n 2 [0 ; 1 10 5 ] . Indonesian J Elec Eng & Comp Sci, V ol. 21, No. 2, September 2021 : 1103 1112 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 r 1105 Ne xt, we describe our chaotic permutation algorithm which is a chaotic e xtension to an im age sc ram- bling procedure, which is introduced in [12]. First of all, the original image scrambling algorithm of [12] is as follo ws. The input to the original image scrambling algorithm is an image with a starting pix el, I P . As sho wn in Fig. 2 (a), suppose that the start position ( S P ) for scrambling of the image is the point (5 ; 3) , pix el with v alue 35 . Based on this point, we di vide the image into sub-blocks Fig. 2 (b). After this partitioning, those sub-blocks are con v erted into linear sequences by the follo wing procedure. F or the S B 1 and S B 2 , we scan them do wn column-wise from left to right and right to left, respecti v ely . The S B 3 and S B 4 are scanned upw ard column-wise from left to right and right to left, respecti v ely . The S B 5 is scanned from left to right in a column-wise f ashion and the S B 6 is scanned from top to bottom, in a ro w-wise f ashion. This is illustrated in Fig. 2 (c). Then, all of the arrays are concatenated as I P , S B 6 , S B 5 , S B 4 , S B 3 , S B 2 and S B 1 to form a single array . Finally , the permuted image is re-constructed from this array in a ro w-wise f ashion Fig. 2 (d). T o generalize this scrambling algorithm and mak e it chaotic, we mak e the follo wing changes: - As we iterat i v ely apply this image scrambling technique for se v eral times, we add an inde x v alue i starting from zero. - The ro ws and the columns of each of the submatrices S B 1 to S B 8 are permuted based on the chaotic v ectors R and C , respecti v ely . T o be precise, at each iterati on i , the ro ws and columns of a sub-matrix S B j of size r ( j ) c ( j ) are permuted with respect to the non-decreasing order of the numbers R [ i : i + r ( j ) ] and R [ i : i + c ( j ) ] , respecti v ely . - Then, if we are in an e v en iteration, we con v ert S B 1 and S B 6 sub-matrices to v ectors by appending their ro ws, and con v ert S B 3 and S B 8 sub-matrices to v ectors by appending their columns. Moreo v er , in e v en iterations, we combine S B 1 through S B 8 and the starting pix el to obtain I P . In odd iterations, we con v ert S B 1 and S B 6 to v ectors column-wise and S B 3 and S B 8 to v ectors ro w-wise. Moreo v er , the I P is constructed by concatenating S B 1 , S B 4 , S B 6 , S B 2 , S B 7 , S B 3 , S B 5 , S B 8 , and the starting pix el. 1 2 9 10 17 18 25 26 3 11 19 27 4 5 6 7 8 12 13 14 15 16 20 21 22 23 24 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 49 50 57 58 43 51 59 44 45 46 47 48 52 53 54 55 56 60 61 62 63 64 5 3 1 7 6 8 4 2 8 1 5 7 3 2 6 4 C( i ) R( i ) S B 1 S B 2 S B 3 S B 4 S B 6 18 17 10 9 2 1 26 25 23 15 7 31 22 14 6 30 20 12 4 28 24 16 8 32 21 13 5 29 63 62 60 64 61 47 46 44 48 45 55 54 52 56 53 58 42 50 57 41 49 27 3 11 19 37 40 36 38 39 59 43 51 34 33 S B 1 S B 2 S B 3 S B 4 S B 5 S B 6 (a) (b) 18 17 10 9 2 1 26 25 23 15 7 31 22 14 6 30 20 12 4 28 24 16 8 32 21 13 5 29 63 62 60 64 61 47 46 44 48 45 55 54 52 56 53 58 42 50 57 41 49 27 3 11 19 37 40 36 38 39 59 43 51 34 33 35 (c) S B 5 Figure 2. Generalized chaotic image scrambling technique. No w , we are ready to describe the proposed encryption algorithm. T o be concrete, the v ariables used in the description of the proposed algorithm are as follo ws: - " denotes the Coupling parameter in the CML - is the logistic mapping parameter in the proposed CML - ` is the Chebyshe v mapping parameter to generate initial sequence x 0 - s 0 is the initial v alue used to generate initial sequence x 0 - m 0 is the number of initial iterations of the Chebyshe v mapping in generating initial sequence to a v oid transient ef fect - m is the length of sequences generated at each le v el - n 0 is the number of initial iterations of the CML in generating sequences to a v oid transient ef fect - n is the total number of sequences of length m to generate by iterating the CML - r is the number of ro ws of the plain image - c is the number of columns of the plain image - K is a symmetric secret k e y used to encrypt the image - X denotes the plain image - Y denotes the encrypted image - h is the hash v alue of the plain image - R is the sequence used to permute ro ws of the plain image Ima g e encryption with a no vel CCML and c haotic ima g e scr ambling tec hnique (Behr ang Chaboki) Evaluation Warning : The document was created with Spire.PDF for Python.
1106 r ISSN: 2502-4752 - C is the sequence used to permute the columns of the plain image - D is the sequence used to dif fuse the pix els in the plain image. The k e y K and the hash v alue of the input image h are gi v en as inputs to the E X T R A C T P A R A M E T E R S algorithm to obtain the required encryption parameters as follo ws: - K 0   K + h mo d 2 256 , - "   K 0 2 256 , - Let K 0 = k 0 32 k 0 31 : : : k 0 2 k 0 1 2 be the binary representation of the modified k e y in 32 Bytes, where k 0 i 2 f 0 ; 1 g 8 for i = 1 ; : : : ; 32 and the plain (encrypted) image be of size r c , - n   max  k 0 31 k 0 29 : : : k 0 3 k 0 1 2 mo d 2 128 ( mo d r ) ; 4 , - n 0   k 0 15 k 0 13 : : : k 0 1 2 , - m   l 4 r c n 2 m + 1 , - m 0   k 0 31 k 0 29 : : : k 0 17 2 + k 0 32 k 0 30 : : : k 0 18 2 , - `   k 0 16 k 0 15 : : : k 0 1 2 + k 0 32 k 0 31 : : : k 0 17 2 , - s 0   ` 2 15 1 , and (10)   ( K 0 mo d 2 64 ) +1 2 64 0 : 43 + 3 : 57 . The E X T R A C T P A R A M E T E R S algorithm is of constant tim e comple xity , gi v en that the k e y length is fix ed. More precisely , i t is linear in the k e y length in bits. W e also use the G E N E R A T E S E Q U E N C E S algorithm to generate chaotic sequences, which is described as follo ws: A chaotic matrix S eq M at of size n 0 + n by m 0 + m chaotic v alues are generated by the proposed CML using the parameters of the Algorithm E X T R A C T - P A R A M E T E R S . The output of the algorithm is a n m sub-matrix of S eq M at by considering its last n ro ws and its last m columns, i.e. S eq M at [ n : ; m :] in Python’ s notation. The G E N E R A T E S E Q U E N C E S algorithm requires ( n 0 + n ) ( m 0 + m ) iterations of the proposed CML, e.g. is of time comple xity O ( m n ) considering e v aluation of the proposed CML requires constant time. The E N C R Y P T I O N algorithm is as follo ws: The hash v alue of the input image X is computed by the SHA-256 algorithm and is denoted as h . The encryption parameters " , , n , n 0 , m , m 0 , ` , and s 0 are e xtracted from the combination of the k e y K and h with the E X T R A C T P A R A M E T E R S algorithm. The algorithm G E N E R A T E S E Q U E N C E S is used to generate three chaotic sequences R , C and D of sizes 2 r , 2 c and ( n 2) m , respecti v ely , as follo ws: (3-a) The first 2 r elements are sorted in a non-decreasing order and their corresponding sorted indices are assigned to the sequence R . (3-b) The ne xt 2 c elements are sorted in a non-decreasing order and their corresponding sorted indices are assigned to the sequence C . (3-c) The ne xt r c elements ar e sorted in a non-decreasing order and their corresponding sorted indices are assigned to the matrix D , which is filled ro w-wise and is of shape r c . The sequences R and C are used to permute the input plain image according to the Step 4 and the sequence D is used to dif fuse the input plain image according to the Step 5. The sequences R and C are used to permute the input image using the chaotic e xtension of the image se gmentation method of [12] by X ( a )   P E R M U T E I M A G E ( X ( a 1) ; i; j ) , for ( R ( i ) ; C ( i )) where R ( i ) and C ( i ) denote the i th element of the sequences R and C , respecti v ely , and a = 1 ; : : : ; min r ; c . Let X ( m ) be the output of thi s step. - The sequence D is used to dif fuse the permuted image X ( m ) by X D   X ( m ) + D mo d 256 . pair Y = ( X D ; h ) is the output of the E N C R Y P T I O N . The first and the second steps of the encryption algorithm can be considered of constant time com- ple xity . In the third step of the encryption algorithm, we need to generate m n chaotic elements from the CML sequence and then, sorting them which requires O (( m n ) log ( m n ) + m n ) operations. In the fourth step, the input image is scrambled for min r ; c steps, where each step requires r c operations. Finally , the last step can be accomplished with r c operations. Therefore, the time comple xity of this algorithm is O (( m n ) log ( m n ) + r c min( r ; c ) + r c + 1) , e.g. it is polynomial in terms of the size of the input image. Finally , an encrypted image can be decrypted by re v ersing the encryption process. More precisely , the D E C R Y P T I O N is as follo ws: Let the input of the algorithm be the pair Y = ( X D ; h ) and the secret k e y K . The encryption parameters " , , n , n 0 , m , m 0 , ` , and s 0 are e xtracted from the combination of the k e y K and h with the E X T R A C T P A R A M E T E R S algorithm. Indonesian J Elec Eng & Comp Sci, V ol. 21, No. 2, September 2021 : 1103 1112 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 r 1107 The chaotic sequences R , C and D are generated with the same procedure as in the third step of the encryption algorithm. The dif fusion is re v ersed by computing Y D = Y D mo d 256 . The permutation is re v ersed by re v ersing each step of the chaotic image scrambling using sequences R and C . It is easy to v erify that the time comple xity of the encryption algorithm and the decryption algorithm are the same, i.e. polynomial in the size of the image. 3. RESUL TS AND DISCUSSION An ef ficient encryption algorithm must demonstrate a proper security performance in e n c ou nt ering dif ferent security attack and could withstand them rigorous ly . T o test the security of our encryption algorithm, we analyzing common f actors, such as k e y space analysis, k e y and plainte xt sensiti vity , histogram analysis of encrypted image, entrop y analysis, and correlation coef ficient analysis. The proposed algorithm is implemented in Python 3.6.7 and the results are obtained on an Intel C orei3- 350 M Processor 2.26 GHz running Ub untu 18.04 64-bit. The running t ime of the proposed algorithm is 1.44 seconds on a v erage for plainte xt images of size 512 512 using 256 -bits k e y with this implementation. Our proposed algorithm is secure ag ainst the brute-force attack, since its k e y space is of size 2 256 . Sensiti vity ag ainst the encryption/decryption k e y is a requirement for a rob ust image encryption algo- rithm. There are three types of k e y sensiti vity: (1) Decryption of an encrypted image with single bit perturbation in the le gitimate k e y , (2) Decryption of an encrypted image with ille g al k e ys, and (3) Encryption of the same image with single bit perturbation in the k e y . It is required that the result of each sensiti vity analysis be dif ferent to each other as much as possible. The results of these tests for the proposed algorithm are reported in T ables 1-(a), 1-(b), and 1-(c), respecti v ely . As it can be observ ed from these tables, the proposed algorithm pro vides acceptable sensiti vity to t he encryption k e y , as all these v alues are v ery close to their ideal v alues. Moreo v er , the encryption sensiti vity of the proposed algorithm in the k e y is much higher t han some recent chaotic image encryption algorithms, as it can be observ ed from T ables 1-(a) to 1-(c). The main purpose of these tests is to ensure the dif fusion propert y of an encryption algorithm, which means that the smallest change in plain image should ha v e huge consequences in cipher image. The Uni- fied A v erage Changing Intensity (U A CI) and the Number of Pix els Change Rate (NPCR) are tw o of the most standard (or another synon ym) tests that are used by researchers for testing the resistance of their encryption algorithm ag ainst the dif ferential attacks (or chosen-plain te xt attacks ).The NPCR is used to calculate the per - centage of dif ference between to image pix el numbers, on the other hand the U A CI is used to measure the a v erage se v erit y of dif ferences between the tw o images. The ideal v alues of NPCR and U A CI based on the e xpectations of [13] are 99 : 6094% and 33 : 4635% , respecti v ely . The NPCR and U A CI tests for tw o images C 1 and C 2 of size M N are defined as the follo wing equations NPCR = P M i =1 P N j =1 D ( i;j ) M N 100% ; and U A CI = P M i =1 P N j =1 j C 1 ( i;j ) C 2 ( i;j ) j 256 M N 100% ; respecti v ely , where D ( i; j ) = 1 if C 1 ( i; j ) = C 2 ( i; j ) , otherwise D ( i; j ) = 0 . In some cases, the NPCR and U A CI cannot accurately detect the visual dif ferences between the plain-image and cipher -image, so to compensate these criteria, we use B A CI (Block A v erage Changing Intensity) test for analyzing our encryption algorithm. This test quantit ati v ely e v aluates the dif ferences between our plain and cipher -images. Let M and P be tw o images of the size r c 3 . Then, the v alue of their PSNR is calculated as PSNR ( M ; P ) = 20 log 255 MSE where MSE ( M ; P ) = 1 r c 3 P r i =1 P c j =1 P 3 k =1 ( M ( i; j ; k ) P ( i; j ; k )) 2 . Note that high numerical dif ferences between the plain and the decrypted image results in lo wer v alues of PSNR. In a dif ferential attack, an attack er slightly changes the plain image and encrypts the original plai n and the modified images with the same k e y . Then, he tries to trace the dif ferences between the tw o encrypted images and use this kno wledge to crack it. Our proposed algorithm pro vides acceptable sensiti vity with respect to dif ferential attack as is illustrated in T able 1-(d). Moreo v er , our proposed algorithm outperforms most of the recent image encryption algorithms, as it is sho wn in T able 1-(d). In image processing and image encryption c o nt e x t , the histogram is a statistical feature of image that sho ws the frequenc y of each pix el intensity in grayscale or e v en color image. F or grayscale images that are presented in this article, distrib ution of 256 dif ferent possible combinations are sho w graphically by image histogram. As depicted in the Figure 3, the uniform distrib ution of cipher image histogram sho ws that our algorithm resistance ag ainst statistical attacks. Ima g e encryption with a no vel CCML and c haotic ima g e scr ambling tec hnique (Behr ang Chaboki) Evaluation Warning : The document was created with Spire.PDF for Python.
1108 r ISSN: 2502-4752 T able 1. Sensiti vity analysis where “—” denotes that the corresponding v alue is not a v ailable. (a) Image Decryption sensiti vity in the secret k e y NPCR U A CI B A CI PSNR Pepper 99.61578 30.69388 0.23172 8.53623 Lena 99.61231 28.98183 0.21558 9.09798 Baboon 99.60728 27.95127 0.20694 9.48089 Ship 99.60918 28.44964 0.20764 9.29461 Ceremon y 99.60678 32.66921 0.25568 7.95964 Cameraman 99.61529 28.85233 0.21541 9.15081 Girl 99.60777 32.39105 0.25241 8.03456 Home 99.60735 27.61322 0.19721 9.61956 A v erage 99.61022 29.70030 0.22282 8.89678 [14] 99.60860 28.69690 0.21382 [15] 99.60890 28.62900 0.21325 [16] 99.60960 28.6321 0.21321 (b) Image Decryption sensiti vi ty in ille g al k e ys NPCR U A CI B A CI PNSR Pepper 99.60403 33.45612 0.26634 7.75197 Lena 99.61967 33.43398 0.26623 7.75487 Baboon 99.61891 33.37835 0.26645 7.76399 Ship 99.63074 33.46454 0.26656 7.75178 Ceremon y 99.60785 33.47861 0.26727 7.74123 Cameraman 99.61967 33.41536 0.26648 7.75820 Girl 99.62425 33.43350 0.26633 7.75333 Home 99.60136 33.44234 0.26623 7.75573 A v erage 99.61581 33.43875 0.26648 7.75388 [17] 63.62070 29.21160 0.34080 [15] 99.60760 33.47100 0.26779 [16] 99.60740 33.46030 0.26777 (c) Image Encryption sensi ti vity in the secret k e y NPCR U A CI B A CI PSNR Pepper 99.61578 33.49226 0.26669 7.74241 Lena 99.61231 33.48726 0.26704 7.73877 Baboon 99.60728 33.45494 0.26674 7.74823 Ship 99.60918 33.43114 0.26643 7.75608 Ceremon y 99.60678 33.51472 0.26706 7.73511 Cameraman 99.61529 33.46900 0.26647 7.74836 Girl 99.60777 33.45926 0.26671 7.74803 Home 99.60735 33.48952 0.26667 7.74308 A v erage 99.61022 33.47476 0.26672 7.74501 [18] 99.3097 33.4553 —– —– [19] 99.60269 33.49419 —– —– [20] 99.418 33.39 —– —– [21] 99.6074 33.4570 —– —- [22] 99.61596 33.43452 —– —– [23] 99.60867 33.49567 —– —– (d) Image Encrypti on of P erturbed Image with Same K e y NPCR U A CI B A CI Pepper 99.60964 33.44334 0.26650 Lena 99.61067 33.46390 0.26670 Baboon 99.61662 33.46825 0.26680 Ship 99.61140 33.43944 0.26657 Ceremon y 99.60907 33.46023 0.26691 Cameraman 99.60968 33.45641 0.26670 Girl 99.60316 33.48295 0.26665 Home 99.61449 33.46452 0.26667 A v erage 99.61059 33.45988 0.26669 [24] 99.61 33.53 [25] 99.61 33.43 [14] 99.6092 33.4668 0.2678 [21] 99.6074 33.4570 [26] 99.61 33.33 [20] 99.61 33.45 In a natural image, the correlation between adjacent pix els is usually high. Due to this high correlation between tw o adjacent pix els, the encryption algorithm could be vulnerable ag ainst statistical attacks. So, an encryption algorithm is as rob ust as possible if it could break do wn this relationship as much as possible. T o measure the correlation of pix els in plain and cipher -images we randomly choose 3 000 pix els form original and corresponding cipher images in the v ertical , horizontal and diagonal direction, and then compute the correlation coef ficients. The results are gi v en in T able 2-(a). As it can be observ ed from thi s data, our proposed algorithm outperforms four out of se v en recent image encryption algorithms in all three directions, and outperforms the rest of them in at least one direction. Thus, it pro vides acceptable security in terms of correlation analysis. The concept of informat ion entrop y is a solution for finding a de gree of randomness and uncertainty in the image. As the information entrop y of an image becomes close to its ideal v alue, it suggests that the infor - mation is distrib uted randomly throughout the image. F or grayscale images, the optimum v alue for information entrop y is 8, so as f ar as our analysis results become closer to this v alue, we can conclude that our encryption algorithm has a better uniform distrib ution of information. The entrop y of a grayscale image P is calculated by the follo wing equations: H ( P ) = P 255 i =0 p ( i ) log 2 p ( i ) ; where p ( i ) is the fraction of pix els with color i in the image P . As it can be v erified in T able 2b, our proposed method pro vides acceptable security in terms of entrop y analysis compared with se v eral recent image encryption algorithms. It is vital for a secure image encryption algorithm to be rob ust ag ainst data loss and noise attacks, since the presence of noise and data loss is quite usual in real w orld scenarios. The proposed algorithm satisfies this requirement as we ha v e tested it by cropping 200 200 random square from an encrypted image and then decrypted it. Moreo v er , we ha v e also added a Salt & Pepper noise of 0 : 1% , 1% and 5% to the encrypted image and decrypted it. As it can be visually observ ed in Figures 4 and 5, the proposed encryption scheme pro vides an acceptable rob ustness ag ainst noise and data loss attacks. Indonesian J Elec Eng & Comp Sci, V ol. 21, No. 2, September 2021 : 1103 1112 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 r 1109 (a) Plainte xt image. (b) Encrypted image. (c) Plainte xt image histogram. (d) Cipherte xt image histogram. Figure 3. Histogram analysis for the proposed algorithm for a sample image. (a) (b) (c) (d) Figure 4. Crop Attack 200 200 . (a) Plainte xt image, (b) Encrypted image, (c) Cropped encrypted image, (d) Decrypted image of the cropped encrypted image. (a) (b) (c) (d) (e) (f) (g) (h) (i) (j) Figure 5. Noise attacks: (a) Plainte xt image, (b) Encrypted image, (c) Salt and Pepper noise of 0 : 01 added to encrypted image, (d) Decryption of encrypted image with Salt and Pepper noise of 0 : 01 added, (e) Salt and Pepper noise of 0 : 1 added to encrypted image, (f) Decryption of encrypted image with Salt and Pepper noise of 0 : 1 added, (g) Salt and Pepper noise of 0 : 5 added to encrypted image, (h) Decryption of encrypted image with Salt and Pepper noise of 0 : 5 added, (i) Gaussian noise of mean 0 and v ariance 0 : 01 added to encrypted image, (j) Decryption of the encrypted image with Gaussian noise of mean 0 and v ariance of 0 : 01 added. Ima g e encryption with a no vel CCML and c haotic ima g e scr ambling tec hnique (Behr ang Chaboki) Evaluation Warning : The document was created with Spire.PDF for Python.
1110 r ISSN: 2502-4752 T able 2. Correlation and entrop y analysis for plain and encrypted images with the proposed algorithm for all images. (a) Correlation analysis Image Correlation direction Horizontal V ertical Diagonal Pepper 0.01403 0.00189 -0.00372 Lena -0.03244 0.00820 0.04202 Baboon 0.02080 0.01370 0.00952 Ship 0.00194 0.01289 -0.01126 Ceremon y 0.00687 0.01787 0.04778 Cameraman 0.00559 0.00523 0.00272 Girl -0.00811 0.01079 -0.00783 Home -0.01582 -0.00224 0.05586 A v erage 0.01320 0.00910 0.02259 [18] -0.09736 -0.07068 0.04844 [19] 0.02424 0.02608 0.02446 [20] 0.0493 0.0172 0.043375 [27] -0.01055 0.0141 0.0081 [22] 0.039886 0.034475 0.001949 [23] 0.07700 -0.07236 -0.06153 [28] 0.0120 0.0917 0.1019 (b) Entrop y for all images Image Entrop y Plain Encrypted Pepper 7.59373 7.99939 Lenna 7.44507 7.99934 Baboon 7.35815 7.99931 Ship 7.19137 7.99933 Ceremon y 6.70385 7.99927 Cameraman 7.15270 7.99934 Girl 7.48451 7.99936 Home 7.20101 7.99929 A v erage 7.39129 7.99933 [19] 5.570216 7.99881 [20] ——– 7.99714 [21] 7.569453 7.989367 [27] 6.501775 7.9972 [22] 7.350343 7.998740 4. CONCLUSION In this paper , we ha v e constructed a no v el chaotic coupled mapped lattice using the chaotic Logistic mapping and the Chebyshe v mapping. W e ha v e also studied the chaotic beha vior of the proposed chaotic mapping in terms of its L yapuno v e xponent, which w as greater t han zero. W e ha v e also generalized an image scrambling method proposed in [12] by making it chaotic and iterati v e. Then, we ha v e combined these tw o primiti v es to encrypt images by generating chaotic sequences from the no v el proposed chaotic mapping and then, applying permutation using the proposed generalized chaotic image scrambling method. T o construct the dif fusion matrix, we ha v e used a sorting approach rather than quantizing the chaotic floating-point v alues. W e ha v e al so studied the security performance of the proposed encryption algorithm concerning brute-force attacks, k e y sensiti vity attacks, dif ferential attack, and statistical analysis. Moreo v er , the rob ustness of the proposed method is tested. The results of all these tests sho wed acceptable security in comparison with se v eral recent image encryption algorithms. REFERENCES [1] X. Chai, Z. Gan, Y . Chen, and Y . Zhang, A visually secure image encryption scheme based on compressi v e sensing, Signal Pr ocessing , v ol. 134, pp. 35–51, 2017. [2] N. Zhou, Y . Hu, L. Gong, and G. Li, “Quantum image encryption scheme with iterati v e generalized arnold transforms and quantum image c ycle shift operations, Quantum Information Pr ocessing , v ol. 16, no. 6, p. Article number: 164, 2017. [3] J. Chen, Z.-l. Zhu, L.-b . Zhang, Y . Zhang, and B.-q. Y ang, “Exploiting self-adapti v e permutation–dif fusion and dna random encoding f o r secure and ef ficient image encryption, Signal Pr ocessing , v ol. 142, pp. 340–353, 2018. [4] Z. Liu, L. Xu, T . Liu, H. Chen, P . Li, C. Lin, and S. Liu, “Color image encryption by using arnold transform and color -blend operation in discrete cosine transform domains, Optics Communications , v ol. 284, no.1, pp. 123–128, 2011. [5] S. Y ou, Y . Lu, W . Zhang, B. Y ang, R. Peng, and S. Zhuang, “Micro-lens array based 3-d color image encryption using the combination of gra vity model and arnold transform, Optics Communications , v ol. 355, pp. 419–426, 2015. [6] S. Putra, H. S. Sheshadri, and V . Lok esha, A na ¨ ıv e visual cryptographic algorithm for the transfer of a compressed medical images, International J ournal of Recent Contri b ut ions fr om Engineering , Science & IT , v ol. 3, no. 4, pp. 26–36, 2015. [7] O. Omoruyi, C. Ok erek e, K. O. Ok okpujie, E. Noma-Osaghae, O. Ok o yeigbo, and S. John, “Ev aluation of the quality of an image encrytion scheme, T elk omnika , v ol. 17, pp. 2968–2974, 2019. [8] R. Ab . Mustaf a, A. A. Maryoosh, D. N. Geor ge, and W . R. Humood, “Iris images encryption based on qr code and chaotic map. T elk omnika , v ol. 18, no. 1, 2020. Indonesian J Elec Eng & Comp Sci, V ol. 21, No. 2, September 2021 : 1103 1112 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 r 1111 [9] A. Shakiba, A no v el randomized one-dimensional chaotic chebyshe v mapping for chosen plainte xt attack secure image encryption with a no v el chaotic breadth first tra v ersal, Multimedia T ools and Applications , 2019. [10] K. Kanek o, “Period-Doubling of Kink-Antikink P atterns, Quasiperiodicity in Antiferro-Lik e Structures and Spatial Intermittenc y in Coupled Logistic Lattice*): T o w ards a Prelude of a “Field Theory of Chaos”, Pr o gr ess of Theor etical Physics , v ol. 72, no. 3, pp. 480–486, 09 1984. [11] K. Briggs, An i mpro v ed method for estimating liapuno v e xponents of chaotic time series, Physics Let- ter s A , v ol. 151, no. 1-2, pp. 27–32, 1990. [12] M. W ang, X. W ang, Y . Zhang, and Z. Gao, A no v el chaotic encryption scheme based on image se gmen- tation and multiple dif fusion models, Optics & Laser T ec hnolo gy , v ol. 108, pp. 558–573, 2018. [13] X. Deng, C. Liao, C. Zhu, and Z. Chen, “Image encryption algorithms based on chaos through dual scrambling of pix el position and bit, J . Commun , v ol. 3, p. 025, 2014. [14] Y . Zhang, “The unified image encryption algorithm based on chaos and cubic s-box, Information Sci- ences , v ol. 450, pp. 361–377, 2018. [15] Z. Eslami and A. Bakhshandeh, An impro v ement o v er an image encryption me thod based on total shuf- fling, Optics Communications , v ol. 286, pp. 51–55, 2013. [16] P . Cheng, H. Y ang, P . W ei, and W . Zhang, A f ast image encryption algorithm based on chaotic map and lookup table, Nonlinear Dynamics , v ol. 79, no. 3, pp. 2121–2131, 2015. [17] K.-w . W ong, A f ast chaotic cryptographic scheme with dynamic look-up table, Physics Letter s A , v ol. 298, no. 4, pp. 238–242, 2002. [18] X. Huang, “Image encryption algorithm using chaotic chebyshe v generator , Nonlinear Dynamics , v ol. 67, no. 4, pp. 2411–2417, 2012. [19] M. J. Rostami, A. Shahba, S. Saryazdi, and H. Nezamabadi-pour , A no v el parallel image encryption with chaotic windo ws based on logistic map, Computer s & Electrical Engineering , v ol. 62, pp. 384–400, 2017. [20] X. W ang, Q. W ang, and Y . Zhang, A f ast image algorithm based on ro ws and columns switch, Nonlinear Dynamics , v ol. 79, no. 2, pp. 1141–1149, 2015. [21] X. W u, K. W ang, X. W ang, and H. Kan, “Lossless chaotic color image cryptosystem based on DN A encryption and entrop y , Nonlinear Dynamics , v ol. 90, no. 2, pp. 855–875, 2017. [22] G. Y e, C. P an, X. Huang, Z. Zhao, and J. He, A chaotic image encryption algorithm based on information entrop y , International J ournal of Bifur cation and Chaos , v ol. 28, no. 01, p. 1850010, 2018. [23] G. Y e and K.-W . W ong, An ef ficient chaotic image encryption algorithm based on a generalized arnold map, Nonlinear dynamics , v ol. 69, no. 4, pp. 2079–2087, 2012. [24] J. W u, X. Liao, and B. Y ang, “Image encryption using 2d h ´ enon-sine map and DN A approach, Signal Pr ocessing , v ol. 153, pp. 11–23, 2018. [25] Q. Y in and C. W ang, A ne w chaotic image encrypt ion scheme using breadth-first search and dynamic dif fusion, International J ournal of Bifur cation and Chaos , v ol. 28, no. 04, p. 1850047, 2018. [26] W . Liu, K. Sun, and C. Zhu, A f ast image encryption algorithm based on chaotic map, Optics and Laser s in Engineering , v ol. 84, pp. 26–36, 2016. [27] L. Xu, Z. Li, J. Li, and W . Hua, A no v el bit-le v el image encryption algorithm based on chaotic maps, Optics and Laser s in Engineering , v ol. 78, pp. 17–25, 2016. [28] N. Zhou, H. Jiang, L. Gong, and X. Xie, “Double-image compression and encryption algorithm based on co-sparse representation and random pix el e xchanging, Optics and Las er s in Engineering , v ol. 110, pp. 72–79, 2018. BIOGRAPHIES OF A UTHORS Behrang Chaboki is a lecturer of Computer Science at the V ali-e-Asr Uni v ersity of Rafsanjan. His research interests include Algorithms and Computational Comple xity . He holds a M.Sc. in Computer Science from Sharif Uni v ersity of T echnology . Ima g e encryption with a no vel CCML and c haotic ima g e scr ambling tec hnique (Behr ang Chaboki) Evaluation Warning : The document was created with Spire.PDF for Python.
1112 r ISSN: 2502-4752 Ali Shakiba is an assistant professor of Computer Science a t the V ali-e-Asr Uni v ersity of Rafsanjan. His research interest inc lude Cryptograph y , Machine Learning and Chaotic Systems. He holds a Ph.D. in Computer Science from Y azd Uni v ersity . Indonesian J Elec Eng & Comp Sci, V ol. 21, No. 2, September 2021 : 1103 1112 Evaluation Warning : The document was created with Spire.PDF for Python.