Compensation of degradation, security, and capacity of LSB substitution methods by a new proposed hybrid n-LSB approach

Computer Science and Information Systems 00(0):0000–0000 https://doi.org/10.2298/CSIS123456789X 3 Compensation of degradation, security, and capacity of LSB substitution methods by a new proposed hybrid nLSB approach 4 Kemal Tütüncü, Özcan Çataltaş* 5 6 7 Faculty of Technology, Selcuk University, 42130 Konya, Turkey {ktutuncu, ozcancataltas}@selcuk.edu.tr 1 2 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Abstract. This study proposes a new hybrid n-LSB (Least Significant Bit) substitution-based image steganography method in the spatial plane. The previously proposed n-LSB substitution method by authors of this paper is combined with the Rivest-Shamir-Adleman (RSA), RC5, and Data Encryption Standard (DES) encryption algorithms to improve the security of the steganography, which is one of the requirements of steganography, and the Lempel-Ziv-Welch (LZW), Arithmetic and Deflate lossless compression algorithms to increase the secret message capacity. Also, embedding was done randomly using a logistic map-based chaos generator to increase the security more. The classical n-LSB substitution method and the proposed hybrid approaches based on the previously proposed nLSB were implemented using different secret messages and cover images. When the results were examined, it has been seen that the proposed hybrid n-LSB approach showed improvement in all three criteria of steganography. The proposed hybrid approach that consists of previously proposed n-LSB, RSA, Deflate, and the logistic map had the best results regarding capacity, security, and imperceptibility. 23 24 Keywords: image steganography, lossless compression, logistic map, data encryption. 25 1. Introduction 26 27 28 29 30 31 32 33 34 35 36 37 38 39 Today, the use of the internet and other technological tools in communication between people is widespread. According to a survey conducted among OECD countries in 2019, internet access on an individual basis has increased from 45.7% to 85.6% between 20052018 [1]. As the use of the internet increases in communication between people, privacy concerns also increase. For this reason, efforts to ensure privacy in communication have increased. Steganography is one of the data hiding sciences that aims to increase confidentiality in communication [2-3]. The primary purpose of steganography is to conceal the existence of a secret message. This purpose is the most crucial feature that distinguishes it from other data hiding sciences. Since a media file containing a secret message will not be attracted by the third party viewing the message, the secret message will not arise. Therefore, researchers' interest in this subject has increased gradually. Unlike other data hiding methods, the message is hidden in another media file called a cover or carrier in steganography. This file type can be text, image, audio, video and, etc. 2 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 Kemal Tütüncü, Özcan Çataltaş The embedded message received by the recipient is converted to the original message by reverse conversion. Since the image is used more in communication between people, the studies have focused on the image file as cover media [4]. Cryptography and watermarking are two other concepts used to provide digital information security. In cryptography, the encrypted output of the encryption algorithm attracts the third person's attention to extract the original message [5]. On the other hand, although both techniques have a data hiding scheme, the intent of watermarking is different from steganography. Steganography aims to conceal the existence of any secret message, while watermarking makes it challenging to remove or manipulate the message. Steganography algorithms can generally be divided into two categories: spatial domain and transform domain [6]. In the spatial domain, the secret message's bits are embedded into the cover image by directly manipulating the pixel values. On the other hand, in the transform domain, a secret message is placed in the frequency coefficients calculated from the cover image's pixel values using some mathematical functions. The methods applied in the spatial domain have less computation and time complexity but are relatively less resistant to some attacks. The algorithms in the spatial domain have a very high embedding capacity with very poor perceptibility. In practice, while designing a steganography algorithm, three main features must be considered carefully: imperceptibility, embedding capacity, and security [7]. The embedding capacity and imperceptibility of the stego image are inversely proportional. As the embedding capacity increases, the quality of the stego image decreases. Therefore, using compression methods before embedding the secret message will increase the capacity of the cover media and reduce the detectability of the secret message. The third feature, security, provides resistance against attacks that is subject to steganalysis. Although steganography's main feature is that it is not suspicious, the message can be obtained in case of possible detection of embedding algorithm. Therefore, encrypting the secret message with known cryptology algorithms before embedding it will increase communication security. Another way to improve security in steganography is to embed the secret message randomly instead of sequentially. For this purpose, the embedding process can be done with the help of numbers generated by random number generators [8]. In literature, pseudo-random generators and chaos-based generators are generally used as random number generators. In this study, we have hybridized the different compression methods to increase the embedding capacity of the n-LSB substitution method we introduced in another study [9] and with different encryption methods to increase security. Additionally, we increased security by using a chaos-based (logistic map) embedding algorithm regarding compressed and encrypted messages. These hybrid approaches are tested in different size messages and different images, and the results were compared. It has been seen that the proposed hybrid system compensated degradation, security, and capacity of classical nLSB based image steganography. The paper is organized as follows: In the second part, the existing studies in the literature are examined. In the third part, classical n-LSB substitution method, data compression methods, data encryption methods, random number generators, and image quality evaluation methods are mentioned. In the fourth section, the n-LSB substitution method [9] and the proposed hybrid methods are explained. The obtained results are shown in the fifth chapter. In the sixth section, the results are interpreted, and suggestions are made about future works. Compensation of degradation, security, and capacity of LSB substitution ... 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 2. 3 Related Works In this study, the proposed hybrid methods have been compared with the classical LSB substitution method as can be seen in the following section. Thus, we will include studies in the literature that modified the LSB substitution method or combined it with compression and encryption methods. In our previous study [9], the classical n-LSB substitution method was improved and a new version of n-LSB was proposed and tested on different images. Obtained stego images were compared with stego images obtained by the classical n-LSB substitution method. The proposed n-LSB method caused an increase of 6.6% in the Peak Signal to Noise Ratio (PSNR) value regarding the classical n-LSB substitution method. In their study, Rajput et al. used RSA cryptography and Spatial Orientation Tree Wavelet (STW) compression methods for hiding a secret message in color and gray-scale images. The secret message was encrypted using the RSA encryption algorithm, then embedded in the cover image compressed by the STW compression algorithm. They tested their method using 8 different cover images and obtained PSNR values ranging from 77.3 dB to 83.9 dB [10]. Chen has proposed a new module-based LSB substitution method. In this method, the repeated bits in the secret message are detected and the repeated bits are coded with a code. He tested his method by hiding 7 different gray-scale images at 256x512 pixel resolution in 2 different gray-scale images at 512x512 pixel resolution and obtained the PSNR values ranging from 34dB to 36dB in the test result [11]. Akhtar and colleagues [12] proposed a new module-based LSB steganography method by developing the algorithm proposed by Chen [11]. They tested their method by hiding 10 different gray-scale images with 256x256 pixel resolution in 2 different gray-scale cover images with 512x512 pixel resolution. They obtained PSNR values ranging from 34dB to 40dB in the test result. According to the classical LSB method, they obtained increases of between 3% and 25% regarding PSNR. At the same time, they also applied the method suggested by Chen and emphasized that they achieved a higher PSNR value in their method. Chikouche combined the classical LSB substitution method with the Advanced Encryption Standard (AES) cryptography method and the Deflate compression method in his work. The LSB substitution method was implemented randomly with a pseudorandom generator rather than sequentially. They embedded 3264-bit data in a color cover image with 512x512 pixel resolution and emphasized that their method is better than according to the security criterion [13]. In their study, Manjula and Shivakumar compressed the message they encrypted with AES and Elliptic Curve Cryptography (ECC) with the LZW algorithm and embedded it with the classical LSB substitution method. 32-bit messages were hidden in different images and the PSNR values ranging from 79dB to 81dB values were obtained. Then the messages with a length ranging from 32 bits to 288 bits were hidden in different images and the PSNR values ranging from 77dB to 81dB were obtained. Also, they stated that they have 2 times security because the message is encrypted twice [14]. Kasapbaşi proposed a new image steganography scheme including chaos-based Huffman encoding algorithm and fractal encryption. Firstly, he calculated the frequency of the alphabets and other characters in a section of Turkish newspaper and encoded them with Huffman encoding. He encoded the compressed text with random numbers generated by the logistic map. The message was embedded in the selected LSBs of the 4 Kemal Tütüncü, Özcan Çataltaş 1 2 3 4 5 6 7 8 cover image. The proposed method was found to be successful in terms of encryption [15]. Rachmawanto et al. proposed a hybrid method consists of the AES cryptology method and classical LSB substitution method. They tested their method and obtained PSNR values ranging from 58 dB to 80 dB [16]. Supriadi Rustad et al. proposed a new image steganography method based on finding an adaptive pattern in inverted LSB steganography. They obtained the PSNR value ranges from 52.49 to 57.45, and the SSIM ranges from 0.9991 to 0.9999 [7]. 9 3. Materials and Methods 10 3.1. LSB Substitution Method 11 12 13 14 15 The basic principle of the LSB substitution method is to replace the LSB of each pixel with the message bit in the order of the cover image [17]. It can be applied to RGB or gray-scale images. The value of each pixel, which consists of 8 bits, 0 to 255, is either increased by 1, decreased by 1, or unchanged. A change of ±1 in the image pixel will not make a big difference on the image. 16 3.2. 17 18 19 20 21 22 23 24 25 According to the compression formats, data compression methods are divided into two categories: lossy compression and lossless compression [18]. If the original data can be recovered without any changes after compressing the data, this type of compression is called lossless compression. Lossless compression methods ensure that the original data is preserved precisely and that no detail is desired to be lost. In the other category of compression algorithms, lossy compression, original data cannot be obtained precisely after the recovery. In this study, LZW, Arithmetic, and Deflate algorithms had been chosen to compress the message before it was hidden. Detailed information about these algorithms is shown below. 26 Lempel-Ziv-Welch 27 28 29 30 31 32 33 34 35 The LZW algorithm is a compression method derived from the LZ78 algorithm [18]. It was discovered in 1984 by Terry A. Welch and introduced in his paper titled “A Technique for High-Performance Data Compression” (1984). There is no preset dictionary in the LZW compression method. Dictionary is created dynamically according to the context to be compressed. For this reason, when the LZW method is used to compress the secret message in steganography, the sender does not need to transmit a dictionary to the recipient. Once the recipient has extracted the compressed message from the stego image, the dictionary will be dynamically created, and the secret message will be obtained. Data Compression Methods Compensation of degradation, security, and capacity of LSB substitution ... 5 1 Arithmetic 2 3 4 5 6 7 8 The primary purpose of arithmetic coding is to assign an interval to each character. Then, this range is assigned a decimal number. The algorithm starts with 0 and 1 intervals. After reading each character in the input data, the interval is divided into subparts as a smaller range than the input character's probability. This sub-range becomes the new range and is partitioned according to the probability of that character. This process is repeated for each input character. When this is done, every floating point in the last interval uniquely represents the input data [18-20]. 9 Deflate 10 11 12 13 14 15 16 17 18 19 Deflate is a popular compression method used in well-known algorithms such as Zip and Gzip. Deflate method is used by many important programs such as PNG image, HTTP protocol, and PDF. The Deflate method is a dictionary-based compression technique based on LZ77 and Huffman coding. There are three different modes in Deflate method. In the first mode, input symbols are subdivided without compression. This mode is used for non-compressible files or when someone wants to partition a file without compression. The second mode is a single-pass compression solution. In this mode, a predetermined coding table is used during coding. This mode is used in real-time applications [21]. The third mode of Deflate is a two-pass compression solution based on the dictionaries produced according to the statistical properties of the input file. 20 3.3. 21 RSA 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 The RSA encryption algorithm was proposed by Ron Rivest, Adi Shamir, and Leonard Adleman in 1977 [22]. The expansion of the RSA consists of the initials of the names of the developers. The RSA algorithm is one of the asymmetric encryption methods. Two different keys are used for encryption and decryption, but these two keys are related to each other. The key used for encryption is a public key and is known by everyone. The key used for decryption is a private key and is only found on the receiving side [23]. The encryption steps of the RSA algorithm are as follows: 1. Two prime numbers, such as p and q, are input parameters. 2. The value of 𝑛 = 𝑝 ∗ 𝑞 is the base value, and 𝜑(𝑛) = (𝑝 − 1) ∗ (𝑞 − 1) Euler value is calculated. 3. The number of 𝑒 (public keys) is selected as 1 < 𝑒 < 𝜑(𝑛) (𝜑(𝑛) is a prime number). 4. d value is selected so that 𝑑 ∗ 𝑒 = 1 𝑚𝑜𝑑(𝜑(𝑛)). This value is a private key. 5. The 𝑐 = 𝑚𝑒 𝑚𝑜𝑑(𝑛) formula encrypts each message character. To extract an encrypted message using the RSA algorithm, the first four steps are applied in the same way, and then the secret message is obtained with the formula 𝑚 = 𝑐 𝑑 𝑚𝑜𝑑(𝑛). Data encryption methods Kemal Tütüncü, Özcan Çataltaş 6 1 RC5 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 The RC5 algorithm is one of the symmetric encryption algorithms. It was introduced by Ron Rivest in 1994. The RC5 algorithm is simple to implement because it uses basic mathematical and logical operators. Furthermore, the variable key length distinguishes RC5 from traditional encryption methods such as Data Encryption Standard (DES). The implementation steps of the RC5 encryption algorithm are presented below [24, 25]: 1. Firstly, define w, r, and Key parameters. 2. Obtain P and Q constants. 39 Data encryption standard (DES) 40 41 42 43 44 45 DES was one of the symmetric encryption methods introduced by the National Institute of Standards & Technology (NIST) in 1976 for all government communications. It has also been used for a long time in bank transactions [26]. In the DES algorithm, the input text is divided into blocks. Each has a 64-bit message. A 64-bit key is required in the encryption process. Eight bits of this key are used as parity bits. Encryption is done in 16 rounds. Each round uses a new key that is the 48-bit length. P=odd(e-2)2w Q=odd(Φ-2)2w 3. Convert Key K byte to words. for i=b-1 to 0 L[i/u] = (L[u/i] << 8) + K[i] 4. Initialize key-independent array, S. S[0] = P for i = 1 to 2(r+1)-1 S[i] = S[i-1] + Q) 5. Mix secret key in the L and S array. i = j = 0 A = B = 0 do 3 * max(t, c) times: A = S[i] = (S[i] + A + B) << 3 B = L[j] = (L[j] + A + B) << (A + B) i = (i + 1) % t j = (j + 1) % c 6. Divide the input text into w-bit blocks (A and B are two of these blocks) and encrypt each block. A = A + S[0] B = B + S[1] for i = 1 to r do: A = ((A ^ B) << B) + S[2 * i] B = ((B ^ A) << A) + S[2 * i + 1] 7. Decrypt using A and B. for i = r down to 1 do: B = ((B - S[2 * i + 1]) >> A) ^ A A = ((A - S[2 * i]) >> B) ^ B B = B - S[1] A = A - S[0] Compensation of degradation, security, and capacity of LSB substitution ... 1 2 3 These keys are obtained using the input key. The block diagram of the DES algorithm is shown in Fig. 1. 4 5 Fig. 1. Block diagram of DES encryption algorithm 6 3.4. 7 8 9 10 11 12 13 14 15 16 17 18 19 7 Chaos generator A random number is a series of numbers or symbols that are not predictable with random luck and do not repeat in a particular pattern. Random number generators have many field-critical presets, such as secure communications, data transmission, and storage. Random number generators are examined in two categories: pseudo-random generators and real random generators. Chaos-based generators are used more extensively than pseudo-random generators because they are real random generators. Since the chaos generators are very sensitive to input parameters, the numbers they will produce constantly are not predictable. For this reason, it is frequently used in information security applications [27]. One of the simple chaos systems widely used and applied in the literature is the logistic map. The formula of the logistic map is as follow: 𝑥𝑘+1 = 𝜇 ∗ 𝑥𝑘 ∗ (1 − 𝑥𝑘 ) (1) Here, 𝑥0 and 𝜇 are input parameters. When 3.57 < 𝜇 ≤ 4, the system goes into a chaotic state and generates random numbers. 20 3.5. Image quality evaluation criteria 21 22 Image evaluation criteria are methods used to learn the amount of change in the cover image. The image evaluation criteria used in this study are explained below. 23 Peak Signal-to-Noise Ratio (PSNR) 24 25 26 27 28 29 PSNR is one of the essential criteria used to evaluate image quality. The PSNR is the ratio of the power of the highest possible power of the cover image to the power of the difference between the cover image and the stego image. A high PSNR value means little distortion in the stego image, while a low PSNR value means more distortion in the stego image [28]. The PSNR value can be calculated using the following formula: 255 𝑃𝑆𝑁𝑅 = 20 ∗ 𝑙𝑜𝑔 (2) √𝑀𝑆𝐸 8 Kemal Tütüncü, Özcan Çataltaş 𝑚 𝑛 2 (𝑆(𝑖, 𝑗) − 𝐼(𝑖, 𝑗)) 𝑀𝑆𝐸 = ∑ ∑ 𝑚∗𝑛 (3) 𝑖=1 𝑗=1 1 Average difference 2 3 4 5 The average difference (AD) metric equals the mean of the sum of the differences between the cover image pixels and the stego image pixels. The low average difference value means that there is less distortion in the stego image. The average difference formula is: 𝑚 𝑛 𝑆(𝑖, 𝑗) − 𝐼(𝑖, 𝑗) 𝐴𝐷 = ∑ ∑ (4) 𝑚∗𝑛 𝑖=1 𝑗=1 6 7 8 9 10 11 12 Universal Image Quality Index (UIQI) UIQI (Universal Image Quality Index) is an index that attempts to model any distortion on the image [29]. These distortions can be in the form of a combination of the following three factors: correlation, luminance distortion, and contrast distortion. The UIQI value is between [-1, 1]. 1 means the images are identical. UIQI formula is shown in (5). 𝜎𝑥𝑦 2𝜎𝑥 𝜎𝑦 2𝑥̅ 𝑦̅ 𝑈𝐼𝑄𝐼 = ∗ ∗ (5) 𝜎𝑥 𝜎𝑦 𝑥̅ 2 + 𝑦̅ 2 𝜎𝑥 2 + 𝜎𝑦 2 Here, 𝜎 denotes standard deviation and 𝑥̅ and 𝑦̅ denote the average of the cover and stego images, respectively. 13 4. Proposed Methods 14 15 16 17 18 The classical 1-LSB, 2-LSB, and 3-LSB substitution methods are the most common methods used in steganography because of their ease of implementation and high capacity. Many different methods developed are built on these methods. However, they need to be improved because it is easy to detect, and third-person can directly access the message when it is detected. 19 4.1. 20 21 22 23 24 25 26 27 The pseudo-code of the application steps of the n-LSB substitution method [9] by the author of this paper is presented in Fig. 2. The proposed n-LSB method reduces the excess change between the cover and stego image pixels while applying the classical n-LSB substitution methods. The proposed n-LSB method has only been applied to 2-LSB and 3-LSB methods because there is no effect on the 1-LSB method and also 4 and more bit substitution methods that are not used in the literature. The 2 and 3-bit implementation of the proposed n-LSB method is described and illustrated below. Proposed n-LSB Method Compensation of degradation, security, and capacity of LSB substitution ... Input: C,S,n Output: S C: Cover image pixels S: Stego image pixels n: Embedding method (n=2, 3) 𝐶(𝑝)𝑖 , 𝑆(𝑝)𝑖 : ith bit of related pixel; i=1,..,8 For each p in every pixel of C,S If 𝐶(𝑝) − 𝑆(𝑝) > 2(𝑛−1) If 𝑆(𝑝)𝑛+1 = 0 𝑆(𝑝)𝑛+1 = 1 Elseif 𝑆(𝑝)𝑛+2 = 0 𝑆(𝑝)𝑛+2 = 1 𝑆(𝑝)𝑛+1 = 0 … Elseif 𝑆(𝑝)8 = 0 𝑆(𝑝)8 = 1 𝑆(𝑝)7 = ⋯ = 𝑆(𝑝)𝑛+1 = 0 End If Elseif 𝐶(𝑝) − 𝑆(𝑝) < −2(𝑛−1) If 𝑆(𝑝)𝑛+1 = 1 𝑆(𝑝)𝑛+1 = 0 Elseif 𝑆(𝑝)𝑛+2 = 1 𝑆(𝑝)𝑛+2 = 0 𝑆(𝑝)𝑛+1 = 1 … Elseif 𝑆(𝑝)8 = 1 𝑆(𝑝)8 = 0 𝑆(𝑝)7 = ⋯ = 𝑆(𝑝)𝑛+1 = 1 End If End If End For each 1 2 Fig. 2. The pseudo-code of the proposed n-LSB substitution method 3 Proposed 2-LSB substitution method 4 5 6 7 8 9 10 11 12 9 When 𝑛 = 2 in the pseudo-code given in Fig. 2, the secret message is embedded in the cover image using the classical 2-LSB substitution method. At the end of the embedding process, the difference between the cover image and the stego image pixels is examined. Since at most two LSBs of the cover image pixels can be changed, the decimal difference will be one of 3, 2, 1, 0, -1, -2, -3. If this difference is -2, -1, 0, 1, or 2, then the pixel of the stego image is left unchanged. In another case, if this difference is 3, all bits from the 3rd LSB of the stego image are examined, the first 0 is converted to 1, and the previous 1s are converted to 0. So the related pixel of the stego image is added to the decimal number 4, the difference between the pixels falls from 3 to -1. If there are no 0's between Kemal Tütüncü, Özcan Çataltaş 10 1 2 3 4 5 6 7 8 the 3rd and 8th bits (most significant bit), the related pixels are left unchanged in the stego image. If the difference is -3, all bits from the 3rd bit of the related pixel are examined, the first 1 encountered is converted to 0, and the previous 0s are converted to 1. Thus, the decimal number 4 is subtracted from the corresponding pixel, and the difference between the pixels falls from -3 to 1. If there is no 1 between the 3rd and 8th bits, then the corresponding pixel is left unchanged in the stego image. With this method, the deterioration of the pixels where the degradation is excessive is reduced [9]. Examples of the implementation of the proposed 2-LSB are shown in Table 1. 9 10 11 Table 1. Proposed 2-LSB substitution example. The changed bits of cover image pixel obtained after embedding with the 2-LSB method and the bits of stego image pixel obtained after applying the proposed compensation method were shown in red and green color, respectively Pixel No Cover image Message to be embedded (2-bit) Stego image Difference New stego image New difference 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 01111001 00100011 10010000 00000001 11000011 11000010 00101110 11111000 00110000 11010100 11100000 00100000 01010001 01101101 11101001 11001101 01110111 01111110 01001101 10000011 01 10 11 10 00 11 01 01 10 11 11 10 01 10 10 01 00 10 10 10 01111001 00100010 10010011 00000010 11000000 11000011 00101101 11111001 00110010 11010111 11100011 00100010 01010001 01101110 11101010 11001101 01110100 01111110 01001110 10000010 0 1 -3 -1 3 -1 1 -1 -2 -3 -3 -2 0 -1 -1 0 3 0 -1 1 01111001 00100010 10001111 00000010 11000100 11000011 00101101 11111001 00110010 11010011 11011111 00100010 01010001 01101110 11101010 11001101 01111000 01111110 01001110 10000010 0 1 1 -1 -1 -1 1 -1 -2 1 1 -2 0 -1 -1 0 -1 0 -1 1 12 13 14 15 16 17 18 19 20 In Table 1, randomly generated 2-bit messages were hidden in the randomly generated 20 pixels cover image pixels consisting of 8 bits by the 2-LSB method. After the embedding process, the differences between the pixels of the cover image and the stego image were examined. If the difference is -3 or 3, the proposed method was applied. As shown in the table, the difference in 5 of 20 pixels is -3 or 3, so the pixels outside these 5 pixels were not changed. The AD before the enhancement was 1.2, but after the enhancement, this difference was reduced to 0.9. Thereby the amount of distortion in the image was reduced. 21 Proposed 3-LSB substitution method 22 23 24 25 When 𝑛 = 3 in the pseudo-code given in Fig. 2, the secret message is embedded in the cover image using the classical 3-LSB substitution method. At the end of the embedding process, the difference between the pixels of the cover image and the stego image is examined. Since at most three LSBs of the cover image can be changed, the difference Compensation of degradation, security, and capacity of LSB substitution ... 11 1 2 3 4 5 6 7 8 9 10 11 12 13 14 will get one of the values from -7 to 7. If the decimal difference between the pixels is -4, -3, -2, -1, 0, 1, 2, 3, or 4, then the pixel of the stego image is left unchanged. If the difference is 5, 6, or 7, all bits from the 4th bit of that pixel are examined, the first 0 value encountered is converted to 1, and the previous 1s are converted to 0. So the related pixel is added to the decimal number 8, and the difference between the pixels falls from 5, 6, 7 to -3, -2, and -1, respectively. If there is no 0 value between the 4th and 8th bits, the related pixel of the stego image is left unchanged. Similarly, if the difference is -5, -6, or -7, all bits of the pixel are examined from the 4th bit, the first 1 value encountered is converted to 0, and the previous 0s are converted to 1. Thus, the decimal number 8 is subtracted from the related pixel, the difference between the pixels falls from -5, -6, -7 to 3, 2, and 1, respectively. If there is no 1 value between the 4th and 8th bits, then the related pixel is left unchanged. With this method, the deterioration of the pixels where the degradation is excessive is reduced [9]. Examples of the implementation of the proposed compensation method for 3-LSB are presented in Table 2. 15 16 17 Table 2. Proposed 3-LSB substitution example. The changed bits of cover image pixel obtained after embedding with the 3-LSB method and the bits of stego image pixel obtained after applying the proposed compensation method were shown in red and green color, respectively 18 19 20 21 22 23 24 25 26 Pixel No Cover image Message to be embedded (3-bit) Stego image Difference New stego image New difference 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 01111001 00100011 10010000 00011111 11000011 11000010 00101110 11111000 00110000 11010100 11100111 00100000 01010001 01101101 11101001 11001101 01110111 01111110 01001111 10000011 111 001 111 010 100 111 111 001 010 110 000 101 101 101 110 100 001 111 001 001 01111111 00100001 10010111 00011010 11000100 11000111 00101111 11111001 00110010 11010110 11100000 00100101 01010101 01101101 11101110 11001100 01110001 01111111 01001001 10000001 -6 2 -7 5 -1 -5 -1 -1 -2 -2 7 -5 -4 0 -5 1 6 -1 6 2 01110111 00100001 10001111 00100010 11000100 10111111 00101111 11111001 00110010 11010110 11101000 00011101 01010101 01101101 11100110 11001100 01111001 01111111 01010001 10000001 2 2 1 -3 -1 3 -1 -1 -2 -2 -1 3 -4 0 3 1 -2 -1 -2 2 In Table 2, randomly generated 3-bit messages were hidden on the randomly generated 20 cover image pixels consisting of 8 bits by the 3-LSB method. After the embedding process, the differences between the pixels of the cover image and the stego image were examined. If the difference is -7, -6, -5, 5, 6, or 7, the proposed method was applied. As shown in the table, the difference in 9 of 20 pixels is -7, -6, -5, 5, 6, or 7, so the pixels outside these 9 pixels were not changed in the stego image. The AD before the enhancement was 3.45, but after the enhancement, this difference was reduced to 1.85. Thereby the amount of distortion in the stego image was reduced. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 12 Kemal Tütüncü, Özcan Çataltaş 4.2. Proposed hybrid-1 With the proposed n-LSB method [9], the degradation of pixels of the stego image has been reduced, which is one of the three basic principles of steganography. However, there has been no change in the other principles of steganography, which are security and capacity. To improve these two criteria, existing compression and encryption methods are combined with the proposed n-LSB method. Compressing the secret message before embedding it in the cover image will reduce the degradation of the stego image as it will reduce the number of secret message bits to be embedded and increase the message length that can be embedded on the cover image. Also, since the 3rd person will not know the used compression algorithm, the secret message will not be directly available, even if the hidden data in the stego image is recovered. For this purpose, the message is compressed using text compression algorithms before being embedded. Three different algorithms, LZW, Arithmetic, and Deflate, have been applied as text compression algorithms and compared among themselves. Embedding the message sequentially in the cover image makes it easier to extract it by the 3rd person. Encrypting the message before embedding it is a big solution, but since the embedding process is sequential, the 3rd person can quickly get the encrypted message. Although cryptography methods are challenging to break, it is not impossible to break. For this reason, the secret message may be passed on to other people. To overcome this problem, the message in steganography is often randomly embedded in the image with various random number generators rather than sequentially. In our proposed hybrid n-LSB approach, random numbers are generated with the logistic map-based chaos generator to overcome this problem, and the message is randomly embedded in the cover image. The 𝑥0 and 𝜇 values of the chaotic generator are used as input parameters. For this reason, these parameters must be transmitted to the recipient to extract the hidden message. 29 30 Fig. 3. The flowchart of the proposed hybrid-1 method 31 32 33 34 35 36 To eliminate the shortcomings of the n-LSB method and obtain a safer embedding algorithm, two different hybrid methods were created. In the first hybrid method, named as proposed hybrid-1, the secret message was compressed using Deflate compression algorithm, and then this compressed secret message was embedded to the cover image randomly using the proposed 2-LSB and 3-LSB substitution method and logistic mapbased chaos generator. The reason for choosing Deflate compression algorithm is being Compensation of degradation, security, and capacity of LSB substitution ... 13 1 2 3 superior to the other two methods in compression ratios according to applications presented in Results Section. The flowchart of the proposed hybrid-1 method was shown in Fig. 3. 4 4.3. 5 6 7 8 9 10 Proposed hybrid-2 Encrypting the message before embedding it in the cover image will make it difficult to reach the secret message, even if the third party can completely get the information embedded in the image. For this purpose, the message is encrypted with different encryption algorithms. RSA, DES, and RC5 algorithms were tested and compared for this purpose. 11 12 Fig. 4. The flowchart of the proposed hybrid-2 method 13 14 15 16 17 18 19 20 21 22 In this hybrid method, named as proposed hybrid-2, in addition to the proposed hybrid1 method, the secret message was encrypted before compressing process using the RSA encryption algorithm. The reasons why the RSA algorithm is chosen are its widespread use and being asymmetric encryption algorithm. Additionally, the RSA encryption algorithm is superior to the other two methods both in the average encryption time and in the file size to be encrypted per second according to applications presented in Results Section. The (𝑝, 𝑞) values required to generate the public and private keys in RSA encryption are used as input parameters. For this reason, for the receiver to reach the secret message, the values p and q must be transmitted to the receiver. The flowchart of the proposed hybrid-2 method was shown in Fig. 4. 23 5. 24 25 26 27 28 29 30 In this section, the results obtained by applying the classical LSB substitution and proposed hybrid n-LSB approaches are evaluated. For this aim, three different text files and four different cover images were used. The cover images used are shown in Fig. 5. These images are RGB 24-bit images with a resolution of “Lena” 225x225, “Mandrill” 512x512, “Cat” 960x603, and “Peppers” 600x600 pixels. These images are in the ".bmp" file format. The methods we propose in this study are independent of these images and can be applied to any desired image without any constraints. The main reasons for Results Kemal Tütüncü, Özcan Çataltaş 14 1 2 3 4 5 6 7 8 9 10 choosing these images as cover are being in different resolutions and well-known in the literature. Three text files with sizes 6.95 kB, 13.59 kB, and 17.13 kB were selected as secret messages. These text files contain the standard English alphabet as well as some special characters (., ?, -, !, “”, , ). When the results are evaluated, a comparison is made only with the classical LSB substitution method, as seen in the literature. The reason is that each of the methods in the literature is tested on different cover images using different messages. There is no common ground between methods proposed by other authors in the literature. a) 11 b) c) Fig. 5. Cover images used to test proposed methods d) 12 5.1. 13 14 15 16 17 In the proposed hybrid methods, the secret message file is compressed using different compression methods before being hidden in the cover image. The new message length and the compression rates resulting from the compression are shown in Table 3. The compression ratio values were calculated according to (6). 18 19 20 21 22 23 24 25 26 Comparison of data compression methods %𝐶𝑜𝑚𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛 𝑅𝑎𝑡𝑖𝑜 = 𝐼𝑛𝑝𝑢𝑡 𝑆𝑖𝑧𝑒 ∗ 100 𝑂𝑢𝑡𝑝𝑢𝑡 𝑆𝑖𝑧𝑒 (6) The original message files with 56980, 111405, and 140364-bit lengths were compressed with the LZW algorithm; their size decreased to 38448, 72969, and 84643bit, respectively. Compression ratios were obtained as 148.2%, 152.67% and 165.83% respectively. It is seen that as the message length increases, the compression ratio increases, and the highest compression ratio is obtained when the Text-3 file is compressed with LZW. The reason is that as the message length increases, the possibility of finding new words added to the dictionary increases. In other words, when there are 2, 3, or more character words added to the dictionary, the probability of encountering these Compensation of degradation, security, and capacity of LSB substitution ... 15 1 2 3 4 5 6 7 8 words increases as the size of the message increases so that the coded words in the dictionary can be used instead of these words. With the arithmetic algorithm, the secret message files, which are 56980, 111405, and 140364-bit length, have been reduced to the size of 35448, 69361, and 88109 bits respectively. Compression ratios were calculated as 160.74%, 160.62% and 159.31%, respectively. It is seen that as the message size increases, the compression ratio decreases, and the highest compression ratio is obtained when the Text-1 file is compressed with the arithmetic algorithm. 9 Table 3. Comparison of compression algorithms Filename Uncompressed size (bit) Text 1 56980 Text 2 111405 Text 3 140364 LZW Arithmetic Compression ratio 35448 160.74% 69361 160.62% 88109 159.31% 38448 148.20% 72969 152.67% 84643 165.83% Deflate 29496 193.18% 56832 196.03% 65744 213.50% 10 11 12 13 14 15 16 17 18 19 20 21 With the Deflate algorithm, the original message files, which are 56980, 111405, and 140364-bits in length, have been reduced to the size of 29496, 56832, and 65744 bits, respectively. Compression ratios were calculated as 193.18%, 196.03% and 213.50%, respectively. Since the Deflate algorithm is a hybrid algorithm consisting of Huffman and LZ77 codes, the highest compression ratios according to other algorithms were obtained with this algorithm. Also, as the message length increased, the compression ratio increased, and the highest compression ratio was obtained when the Text-3 file was compressed. Since the highest compression ratio between LZW, Arithmetic, and Deflate algorithms is obtained by Deflate method, it is used as a compression method in the proposed hybrid1 and hybrid-2 algorithms. 22 5.2. 23 24 25 26 Determining the security and success of encryption algorithms is a complex process. To compare such algorithms on a common basis, encryption times are generally used. Therefore, the encryption times of the RSA, RC5, and DES encryption algorithms used in this study were calculated using texts of different lengths and are shown in Table 4. 27 Table 4. Comparison of encryption times of RSA, RC5, and DES Comparison of data encryption methods Text Size Text-1 (7122 byte) Text-2 (13925 byte) Text-3 (17545 byte) Average Bytes/sec RSA RC5 DES 10.23 24.91 21 20.05 48.29 40.05 24.95 60.97 50.42 18.41 698.75 44.72 287.65 37.15 346.27 16 Kemal Tütüncü, Özcan Çataltaş 1 2 3 4 When Table 4 is examined, it can be seen that the RSA encryption algorithm is superior to other methods both in the average encryption time and in the file size to be encrypted per second. Therefore, RSA was chosen as the encryption method in the proposed hybrid2 algorithm. 5 5.3. Test results 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Three different secret message files were embedded in 4 different cover images using the classical LSB substitution method, the proposed n-LSB method [9], and proposed hybrid methods. The stego images were compared with the cover images by using image comparison criteria, and the results are shown in the sub-sections. The following algorithms are used for embedding: • Classical 1-LSB, 2-LSB, and 3-LSB methods • Proposed n-LSB method (consist of proposed 2-LSB or 3-LSB methods) [9] • Proposed hybrid-1 (consist of proposed 2-LSB or 3-LSB method combined with Deflate compression algorithm and logistic map-based random embedding method) • Proposed hybrid-2 (consist of proposed 2-LSB or 3-LSB method combined with Deflate compression algorithm, RSA encryption algorithm, and logistic map-based random embedding method) The input parameters used during embedding are: • RSA encryption: 𝑝 = 3, 𝑞 = 41. • Logistic map: 𝑥0 = 0.675 and 𝜇 = 3.9763. 21 PSNR 22 23 24 25 26 27 28 29 30 31 32 33 34 The cover and stego images were submitted to the PSNR test and the obtained results are shown in Appendix. Since the PSNR value is the ratio of the peak signal to the noise in the image, the higher PSNR value means that the image degradation is less. When the obtained results are examined, it is seen that the PSNR values of the proposed 2-LSB and 3-LSB methods are higher than the classical 2-LSB and 3-LSB methods. It can be said that the proposed hybrid-1 is better than the proposed n-LSB method and proposed hybrid-2. The second highest PSNR value was obtained by embedding Text-1 in Image4 by the proposed hybrid-1 2-LSB method with 64.86235 dB and comes after the classical 1-LSB method. Also, the highest increase was 10.8426% (from 42.96577 dB to 47.62437 dB) when the Text-3 file was embedded to Image-1 by proposed hybrid-1 3-LSB compared to the classical 3-LSB method. It can be said that the highest PSNR increase can be achieved when the high-size message file is embedded into the low-resolution image. 35 Average Difference 36 37 38 39 The stego images and cover images obtained after applying the proposed and classical LSB substitution methods are compared according to the AD criterion, and the obtained results are presented in Appendix. The average difference is equal to the average of differences between the cover and stego image pixels. Since the stego image is desired to Compensation of degradation, security, and capacity of LSB substitution ... 17 1 2 3 4 5 6 be similar to the cover image, it is expected that the average difference value is small. When the obtained results are examined, embedding Text-1 message on image-4 is obtained with the smallest mean difference value of 0.042 by the proposed hybrid-1 2LSB algorithm. It is also seen that the proposed hybrid-1 is superior to other classical and proposed methods. However, as the resolution of the image increases or the length of the secret message decreases, the average difference value decreases. 7 UIQI 8 9 10 11 12 13 14 15 16 17 18 19 The stego images and cover images obtained after applying the proposed and classical LSB substitution method were compared according to the UIQI criterion, and the obtained results are shown in Appendix. It is preferable to have a high UIQI value because the difference between the stego image and the cover image is desired to be small. The classical 1-LSB method is the least corrupted method because it only changes the last pixels of the cover image. When we compare the proposed n-LSB method with the classical 2-LSB and 3-LSB algorithms, the proposed n-LSB method yields higher UIQI values. Additionally, the results obtained by the proposed hybrid-1 are compatible with the results obtained by the proposed n-LSB. However, since the UIQI value consists of a combination of correlation, luminance distortion, and contrast distortion, the results show differences in different cover images. Therefore, it is not possible to make a clear conclusion about which method is superior according to the UIQI value. 20 5.4. 21 22 23 24 25 26 27 28 29 30 31 32 33 34 Capacity can be defined as the maximum amount of secret messages hidden in the cover image. Thus, it is essential to check the capacity of images when steganography methods are compared. The effect of data compression on stego image capacity is shown in Table 5. The capacity values shown here are estimated values calculated from the compression ratios calculated in Table 3. Besides, since the application of encryption algorithms and other embedding methods such as proposed n-LSB and classical n-LSB do not affect the embedding capacity, only the classical 3-LSB method was used as an embedding method. According to Table 5, the highest capacity increase was achieved by the Deflate algorithm. With this algorithm, the capacity of image-4 was increased from 635.98 kB to 1357.81 kB, a 113.5% increase was obtained. Furthermore, according to the results shown in Table 3, it was obtained that the compression ratio increased as the message size increased. Accordingly, it is expected that with the LZW and the Deflate algorithm, the cover images will have a larger message capacity than the values shown in Table 5. 35 Table 5. Capacity test results Capacity test Image No Image 1 Image 2 Image 3 Image 4 Uncompressed Capacity (kB) 55.62 288.00 395.51 635.98 Compressed Capacity (Expected) (kB) LZW Arithmetic Deflate 92.23 89.40 118.75 477.59 462.93 614.88 655.87 635.74 844.41 1054.64 1022.27 1357.81 18 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 6. Kemal Tütüncü, Özcan Çataltaş Conclusion and Discussion In this paper, steganography methods which are one of the information security methods are examined, and a new hybrid method in image steganography is proposed. This method, which we propose, is based on the proposed n-LSB substitution method of the authors of this paper and tries to reduce the pixel differences between cover and stego image. In this regard, improvement in the perceptibility criterion, one of the three main criteria of steganography, has been achieved and confirmed with an implemented test. Since the proposed n-LSB method is based on the LSB substitution method, the third party can extract the secret message easily. To solve this problem, instead of embedding the message bits in sequence, they are randomly embedded using a chaos-based random number generator. To increase the security a step ahead, RSA, RC5, and DES encryption algorithms are used to encrypt the secret message before being embedded. Then, data compression methods were combined with the proposed n-LSB method to provide improvement in both the capacity criterion and compensating for the increase in data size that would result in the use of encryption algorithms. Three compression methods, LZW, Arithmetic, and Deflate, were applied. The best compression ratio was obtained by the Deflate algorithm. For this reason, the secret message was compressed with the Deflate algorithm before being hidden in the cover image. These proposed hybrid methods based on the proposed n-LSB method have been tested by hiding three message files in different sizes in 4 cover images with different resolutions and sizes. The highest PSNR value was obtained as 64.86 dB with proposed hybrid-1 (2LSB), and the highest PSNR increase rate was 10.84% with proposed hybrid-1 (3-LSB) when the stego images and the cover images were compared according to image quality evaluation criteria. PSNR values were higher in all the different combinations of the proposed n-LSB method than in the classical LSB method. Moreover, the use of the Deflate compression algorithm in the proposed hybrid-1 method resulted in an increase of 113.5% in the embedding capacities of the cover images. Thanks to the proposed hybrid methods, the shortcomings in using the n-LSB method have been eliminated, and more reliable methods have been obtained for data hiding. The proposed n-LSB and hybrid methods can be used regardless of the message and the cover image, as long as the secret message size does not exceed the capacity of the cover image. The authors of this paper study the effects of the application of the combination of different compression and encryption algorithms with the proposed n-LSB method to different color spaces. Furthermore, the authors think that investigating the applicability of the proposed methods in the frequency domain will be a good research step. Compensation of degradation, security, and capacity of LSB substitution ... 1 Appendix 2 Table 6. Test results obtained by applying proposed methods Metric Image No PSNR Image 1 Image 2 Image 3 Image 4 Image 1 AD Image 2 Image 3 Image 4 Image 1 UIQI Image 2 Image 3 Image 4 3 4 5 6 7 8 9 10 11 Text No Text-1 Text-2 Text-3 Text-1 Text-2 Text-3 Text-1 Text-2 Text-3 Text-1 Text-2 Text-3 Text-1 Text-2 Text-3 Text-1 Text-2 Text-3 Text-1 Text-2 Text-3 Text-1 Text-2 Text-3 Text-1 Text-2 Text-3 Text-1 Text-2 Text-3 Text-1 Text-2 Text-3 Text-1 Text-2 Text-3 Classical 1-LSB 55.37 52.46 51.46 62.48 59.61 58.60 63.83 60.93 59.96 65.94 63.06 62.07 0.0625 0.1226 0.1544 0.0121 0.0236 0.0298 0.0089 0.0174 0.0218 0.0055 0.0107 0.0134 0.9914 0.9867 0.9847 1 0.9999 0.9999 0.9994 0.9999 0.9971 0.9989 0.9972 0.9961 2-LSB 51.30 48.41 47.39 58.51 55.58 54.59 59.82 56.91 55.89 61.94 59.03 58.04 0.0793 0.1548 0.1953 0.0152 0.0297 0.0375 0.0112 0.0218 0.0275 0.0069 0.0134 0.0169 0.9789 0.9690 0.9629 0.9999 0.9995 0.9994 0.9940 0.9991 0.9962 0.9945 0.9922 0.9962 3-LSB 46.85 43.98 42.97 53.99 51.10 50.14 55.36 52.44 51.48 57.46 54.49 53.47 0.1108 0.2155 0.2716 0.0214 0.0417 0.0522 0.0156 0.0305 0.0382 0.0096 0.0190 0.0240 0.9645 0.9441 0.9357 0.9999 0.9998 0.9997 0.9985 0.9998 0.9945 0.9989 0.9969 0.9958 Proposed nLSB 2-LSB 3-LSB 53.62 49.69 50.69 46.82 49.68 45.83 60.73 56.86 57.86 53.97 56.85 52.98 62.06 58.10 59.09 55.19 58.03 54.13 64.09 60.08 61.17 57.12 60.17 56.09 0.0627 0.0840 0.1228 0.1634 0.1548 0.2055 0.0121 0.0162 0.0236 0.0316 0.0298 0.0396 0.0089 0.0119 0.0174 0.0232 0.0220 0.0294 0.0055 0.0074 0.0108 0.0146 0.0136 0.0185 0.9859 0.9750 0.9791 0.9612 0.9750 0.9560 1 1 0.9999 0.9999 0.9999 0.9999 0.9994 0.9991 0.9999 0.9999 0.9971 0.9962 0.9991 0.9991 0.9976 0.9976 0.9968 0.9968 Proposed Hybrid-1 2-LSB 3-LSB 54.74 50.42 52.21 48.06 51.72 47.62 61.51 57.12 58.74 54.35 58.14 53.78 62.87 58.44 60.04 55.60 59.46 55.01 64.86 60.42 62.12 57.63 61.49 57.05 0.0443 0.0666 0.0807 0.1167 0.0907 0.1298 0.0092 0.0141 0.0175 0.0267 0.0200 0.0305 0.0067 0.0103 0.0129 0.0199 0.0148 0.0228 0.0042 0.0066 0.0080 0.0125 0.0092 0.0143 0.9908 0.9786 0.9844 0.9683 0.9829 0.9662 0.9998 0.9997 0.9996 0.9994 0.9996 0.9993 0.9949 0.9938 0.9996 0.9994 0.9992 0.9885 0.9961 0.9933 0.9937 0.9894 0.9928 0.9885 19 Proposed Hybrid-2 2-LSB 3-LSB 52.79 48.57 50.52 46.49 50.04 46.10 59.37 55.00 56.60 52.26 55.93 51.57 60.74 56.24 57.83 53.45 57.14 52.73 62.77 58.28 59.90 55.46 59.21 54.74 0.0703 0.1030 0.1224 0.1708 0.1379 0.1891 0.0151 0.0230 0.0288 0.0435 0.0336 0.0509 0.0110 0.0171 0.0215 0.0327 0.0252 0.0385 0.0069 0.0107 0.0134 0.0207 0.0157 0.0243 0.9863 0.9708 0.9787 0.9595 0.9771 0.9571 0.9997 0.9995 0.9994 0.9991 0.9993 0.9989 0.9927 0.9904 0.9993 0.9990 0.9870 0.9825 0.9940 0.9869 0.9909 0.9852 0.9898 0.9840 References 1. OECD. Internet access (indicator). Link 2. Luo X., Wang D., Wang P., Liu F.: A review on blind detection for image steganography. Signal Processing. 88(9). 2138-2157. Link 3. Chen K., Lin C., Zhong S., Guo L.: A parallel SRM feature extraction algorithm for steganalysis based on GPU architecture. Computer Science and Information Systems. 12 (4). 1345-1359. Link 4. Karakus S., Avci E.: A new image steganography method with optimum pixel similarity for data hiding in medical images. Medical Hypotheses. 139. Link 20 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 Kemal Tütüncü, Özcan Çataltaş 5. Tian Q., Han D., Jiang Y.: Hierarchical Authority Based Weighted Attribute Encryption Scheme. Computer Science and Information Systems. 16(3). 797-813. Link 6. Sharafi J., Khedmati Y., Shabani M.M.: Image steganography based on a new hybrid chaos map and discrete transforms. Optik. 226(2). Link 7. Rustad S., Setiadi D., Syukur A., Andono P.: Inverted LSB image steganography using adaptive pattern to improve imperceptibility. Journal of King Saud University - Computer and Information Sciences. Early Access. Link 8. Roy R., Sarkar A., Changder S.: Chaos based Edge Adaptive Image Steganography. Procedia Technology. 10. 138-146. Link 9. Cataltas O., Tutuncu K.: Improvement Of LSB Based Image Steganography. International Journal Of Electrical, Electronics And Data Communication. 5 (2017) 1-5. Link 10. Rajput V., Tiwari S.K., Gupta R.: An enhanced image security using improved RSA cryptography and spatial orientation tree compression method. 2016 International Conference on Signal Processing, Communication, Power and Embedded System (SCOPES), 2016. p. 327-31. Link 11. Chen S.K.: A module-based LSB substitution method with lossless secret data compression. Comput Stand Inter. 33 (2011) 367-71. Link 12. Akhtar N., Ahamad V., Javed H.: A compressed LSB steganography method. 2017 3rd International Conference on Computational Intelligence & Communication Technology (CICT), 2017. p. 1-7. Link 13. Chikouche S.L., Chikouche N.: An improved approach for lsb-based image steganography using AES algorithm. 2017 5th International Conference on Electrical Engineering Boumerdes (ICEE-B), 2017. p. 1-6. Link 14. Manjula Y., Shivakumar K.B.: Enhanced secure image steganography using double encryption algorithms. 2016 3rd International Conference on Computing for Sustainable Global Development (INDIACom), 2016. p. 705-8. Link 15. Kasapbaşi M.C.: A New Chaotic Image Steganography Technique Based on Huffman Compression of Turkish Texts and Fractal Encryption With Post-Quantum Security. IEEE Access. 7 (2019) 148495-510. Link 16. Rachmawanto E.H., Amin R.S., Setiadi D.I.M., Sari C.A.: A Performance Analysis StegoCrypt Algorithm based on LSB-AES 128 bit in Various Image Size. 2017 International Seminar on Application for Technology of Information and Communication (Isemantic). (2017) 16-21. Link 17. Jain N., Meshram S., Dubey S.: Image Steganography Using LSB and Edge – Detection Technique. International Journal of Soft Computing and Engineering. 2 (2012) 217-22. Link 18. Shanmugasundaram S., Lourdusamy R.: A Comparative Study Of Text Compression Algorithms. International Journal of Wisdom Based Computing. 1 (2011) 68-76. Link 19. Langdon G.G., Rissanen J.: Compression of Black-White Images with Arithmetic Coding. IEEE T Commun. 29 (1981) 858-67. Link 20. Langdon G.G.: An Introduction to Arithmetic Coding. IBM J Res Dev. 28 (1984) 135-49. Link 21. Tahghighi M., Mousavi M., Khadivi P.: Hardware implementation of a novel adaptive version of Deflate compression algorithm. 2010 18th Iranian Conference on Electrical Engineering, 2010. p. 566-9. Link 22. Rivest R.L., Shamir A., Adleman L.: A method for obtaining digital signatures ad public-key cryptosystems. Commun ACM. 21 (1977) 1-15. Link 23. Xin Z., Xiaofei T.: Research and implementation of RSA algorithm for encryption and decryption. Proceedings of 2011 6th International Forum on Strategic Technology, 2011. p. 1118 21. Link 25. Rivest R.L.: The RC5 encryption algorithm. in: B. Preneel, (Ed.). Fast Software Encryption. Springer Berlin Heidelberg: Berlin, Heidelberg, 1995. p. 86-96. Link 25. Shahzadi R., Anwar S.M., Qamar F., Ali M., Rodrigues J.J.P.C.: Chaos Based Enhanced RC5 Algorithm for Security and Integrity of Clinical Images in Remote Health Monitoring. IEEE Access. 7 (2019) 52858-70. Link Compensation of degradation, security, and capacity of LSB substitution ... 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 26. McLoone M., McCanny J.V.: High-performance FPGA implementation of DES using a novel method for implementing the key schedule. IET Proceedings - Circuits, Devices and Systems. 150 (2003) 373. Link 27. Patidar V., Sud K.K.: A Pseudo Random Bit Generator Based on Chaotic Logistic Map and its Statistical Testing. Informatica. 33 (2009) 441-52. Link 28. Kadhim I.J., Premaratne P., Vial P.J., Halloran B.: Comprehensive survey of image steganography: Techniques, Evaluations, and trends in future research. Neurocomputing. 335 (2019) 299-326. Link 29. Zhou W., Bovik A.C.: A universal image quality index. IEEE Signal Proc Let. 9 (2002) 81-4. Link Kemal TÜTÜNCÜ was born in Konya, Turkey, in 1975. He received the master’s degree from Free University of Brussel, Belgium. He received Ph.D. degrees from Selcuk University, Turkey. His research interest includes cryptology, information security, natural language processing, and artificial intelligence. Özcan ÇATALTAŞ was born in Konya, Turkey, in 1992. He received the master’s degree from Selcuk University, Turkey. His research interest includes information security and artificial intelligence.