A High Security Digital Image Steganography Based On DWT Using Arnold Transformation with RC4
Jagjot Singh1, Shamandeep Singh2
Department of Computer Science & Engineering,
Chandigarh University, Gharuan, Punjab, India
Abstract - In Steganography, the secret message is embedded in the cover data to form a stego-message in a way that the existence of any embedded information is undetectable, robust, and secure. The data that is used for cover may be in the audio, video, image, or text form. In this paper, a hybrid method for image Steganography, using the combination of RC4 encryption algorithm and Arnold transformation is proposed to improve the security of the stego image without compromising the undetectability of the stego message. Arnold Transform is executed to scramble the secret image with a key. RC4 is used to encrypt the secret image using second key for enhanced security. DWT is implemented to work in the transform domain of the images. Alpha Blending is executed to fuse the secret and cover images. The performance of the proposed method is measured by calculating and comparing various parameters of cover and stego image. The results show that the Security is increased manifolds by using two keys and two-levels of encryption without compromising with stego-image quality.
Keywords – Steganography, Arnold Transform, RC4, Alpha blending, DWT.
INTRODUCTION
The word “Steganography” came from two Greek words: “Stegano” means “covered” and “graphy” means “writing” [1]. Steganography is the science that includes communicating secret information through an appropriate carrier. It is natural that if any doubtful feature is visible or if it is known that the file is carrying any secret data, the attack would be evident, thus the goal of the Steganography is to hide the very existence of the embedded data.
Steganography’s purpose in security is to supplement cryptography, not replace it and moreover, if a hidden message is ciphered, it must be deciphered if hacked, which adds another layer of protection [27].
Watermarking and cryptography are the cousin techniques to Steganography. The main factor that separate Steganography from Watermarking and cryptography is that the Steganography makes the secret message undetectable.
In cryptography, the existence of the secret data is detectable but is indecipherable [14]. Unlike cryptography, Steganography does not attract public attention. Thus, cryptography may or may not be a part of Steganography i.e. image can be encrypted before embedding. This gives Steganography an added benefit over cryptography.
On the sender side, cover image and secret image are embedded using an embedding process resulting in stego image. The receiver side is required to implement the inverse of the same using cover and stego image. Steganography is used for confidential communication, secret data storage, communicating, and nation’s defense strategies.
1. RELATED WORK
Prabakaran.G. et.al., [1] in 2012 proposed a method of secure image Steganography using DWT, Arnold transform, alpha blending. This method of Steganography is highly secured as the image is scrambled first and then is embedded. Chen.P et al., [2] have proposed that the secret message should be embedded into the high frequency co-efficient of the wavelet transform while leaving the lower frequency co-efficient unaltered. Ge XIuhui et al., [3] have done some research on the applications of immune digital water marking algorithm and found out that wavelet transforms have certain robustness against some image processing methods. Manjunatha Reddy H.S., et al., [4] proposed an algorithm where only the approximation band of the secret image is considered and is fused based on alpha and beta which increases the capacity of the cover image.
Kumar.V et al., [5] in 2010 observed the effect of embedding the secret image in all the sub bands (approximation, detail) of the cover image. PSNR is highest when secret message is embedded into diagonal detail sub-band of the cover image. Po-Yueh Chen et al., [6] proposed a frequency domain method. Firstly, cover image is transformed into sub bands. The secret image is embedded into approximation sub band of the cover image. Shikha Sharda et al., [7] did propose a method based on DWT and Arnold transform along with a private key to provide high security. Yih-Kai Lin [8] proposed a lossless Steganography scheme that embeds secret image into DCT coefficients of the cover image.
A.Nag et al., [19] proposed a novel approach for image Steganography using DWT and Huffman encoding. Boora.M et al., [9] included Arnold transform in Steganography to obtain scrambled secret image. The method is highly secured with satisfactory perceptual invisibility. Zhenjun Tang et al., [10] proposed the combination of Arnold transform with three other random techniques for image encryption. The scheme thus created is robust and secure. Prabakaran.G et al., [11] proposed a Steganography method of using dual wavelets (DWT & IWT). It hides smaller size secret image into larger size cover image using different combinations of DWT and IWT. Arnold transform is used to scramble secret image. Ghasemi et al., [12] proposed to embed data in IWT sub bands by using genetic algorithm based mapping function and increase the hiding capacity with low distortions. Anil Kumar Katharotiya et al., [13] did a comparative analysis of two different image compression techniques based on DCT and DWT. It tells us that information loss in DCT is more than information loss in DWT. Hence, DWT is better than DCT in quality but DCT is better than DWT in terms of time consumption.
Shraddha Soni et al., [14] analyzed AES and DES encryption techniques for their strengths and weakness. H.S. Manjunatha Reddy [15] proposed a method for high capacity and high security Steganography using DWT, alpha blending. Capacity is increased by considering only the approximation band of the secret image. Agniswar dutta [16] proposed that the secret message is first encrypted and then is embedded in the spatial domain of the cover image. Hameed A. Younis et al., [17] proposed a modified RC4 algorithm and used it with Wavelet transform.
Issa A.Abed [18] used RC4 encryption algorithm with wavelet transform. Chetan K. Solanki et al., [20] compares the image fusion methods based on pixels and wavelets. The results show that the wavelet based fusion techniques gives better visual output than the pixel based techniques. Sasidharan.S et al., [21] proposed a secure encryption method using DWT with RC4 stream cipher. This method provides a high level of security by scrambling the rest of the image using some scrambling algorithm. Adnan M.A.Brifcani et al., [22] proposed a technique based on cryptography and Steganography algorithms. The proposed technique uses RSA encryption algorithm to increase security with IWT. Khalid Hamdnaalla et al., [23] proposed a cryptographic method in which image is first scrambled using Arnold transform and then is encrypted using RC4 stream cipher for two level security. Anoop.S et al., [24] compared the performance of stream ciphers like RC4 in combination to DWT, DCT, and spatial domain. The results show that PSNR of DWT case is higher and is resistant to statistical attacks.
R.Nivedhitha et al., [25] proposed a combination of cryptography and Steganography by encrypting the secret image with DES algorithm and then embedding it in the spatial domain of the cover image. Jawahar Thakur et al., [26] did performance analysis of three encryption algorithms namely, AES, DES, blowfish. The results show that blowfish is the best encryption algorithm. AES requires high processing power, so gives poor performance.
2. PROPOSED METHOD
In this proposed method, we have used the combination of Arnold Transform and RC4 with DWT to provide high security. Two keys, one with Arnold Transform and second with RC4 are used. In the encoding process, The Arnold transform results in scrambled secret image. This increases the robustness as well as security of the algorithm. RC4 is then used to encrypt the scrambled image to further increase the security. DWT is then applied to the cover and scrambled secret image and the corresponding sub bands are added by applying alpha blending. IDWT is then applied to get stego image. The decoding process is reverse of the encoding process.
3.1Discrete Wavelet Transform
The wavelet based transform gives the time-frequency representation of the signal simultaneously. In DWT, the signal is passed through the low pass and high pass filters [28]. The outcome of low pass filter represents the low resolution version of the original image and is called approximation coefficient (A), and that of high pass filter represents the residual information of the original image and is called detail coefficient.
In 2D image decomposition at level-1, the image is passed through low and high pass filters first in the horizontal direction and then in the vertical direction. The approximation sub band contains the major information of the image. The detail sub bands include horizontal (H), vertical (V) and diagonal (D) sub bands. For level-2 decomposition, the approximation band from level 1 is decomposed further in the same manner.
(a) /
(b)
Figure 3.1. (a) Level-1 DWT, (b) Level-2 DWT
3.2Arnold Transform
Arnold transformation is a scrambling algorithm also known as cat face transform [9] proposed by V. J. Arnold. Image matrix is given as input to this algorithm, and we get a chaotic or scrambled matrix as output. Image matrix should be a square matrix of N*N size. It is defined as
X'Y'= 1112 XY mod(N) (1)
Here, N is rows of square matrix, X, Y are pixel coordinates of original image and X’, Y’ are pixel coordinates of scrambled image
3.3Rivest Cipher 4
RC4 is a stream cipher encryption algorithm i.e. it encrypts data one bit or one byte at a time, following private key encryption i.e. one key is used for both encryption and decryption [23]. Because RC4 is not a block cipher, therefore it is quite fast and use far less code compared to the block ciphers. Although, RC4 has many weaknesses, still many applications that require faster processing make use of it. RC4 algorithm has 2 parts - Key Scheduling Algorithm does initialize the permutation in the array. Key length is defined as the number of bits in the key and may range from 1 to 256 bits. After initializing the array to identity permutation, it is processed for 256 iterations and then mixed with the key. Through KSA a key stream is generated.
Pseudo Random Generator Algorithm states that Number of iterations are equal to number of pixels in the image (N*N). PRGA modifies the key stream for each iteration. Then, for encryption, the modified key stream is XORed with the message to produce a ciphered message. For decryption, the modified key stream is XORed with the ciphered message to get original message.
3.4Alpha Blending
Alpha Blending [9] is a method of fusing two images together to form a stego image. In this algorithm the components of the cover image and secret image are first multiplied by a constant value or scaling factor, 0.0 ≤ α ≤1.0 and then are added. The value of alpha determines the mixing level of both the images. The equations for implementing alpha blending are
Fused image = 1-α*Cp+ α*Sp(2)
Recovered image = St- Cp * 1- αα(3)
Here, Cp is cover image, Sp is Secret image, St is Stego image.
3.5Implementation of Proposed method
This topic describes the decoding and encoding processes clearly.
3.5.1Encoding Process
Figure 3.2. Encoding process of proposed Steganography
The encoding process includes DWT, Arnold Transform, RC4, Alpha blending, IDWT. The combination of RC4 and Arnold transform along with two keys gives more security in the proposed method. The stego image produced can be transferred to any communication media. The schematic diagram of encoding process is shown in figure 3.2.
Algorithm for encoding process
1. Select cover Image, C and Secret image, S both of size N*N.
2. On S, apply Arnold Transform with a private key, K1 to get scrambled image, Sc.
3. Encrypt Sc with RC4 algorithm with second key, K2 and get ciphered image, Cip.
4. Perform 2-D DWT at level 1 for C and Cip. We will get approximation and detail sub bands of both C and Cip.
5. Blend these sub bands using alpha blending choosing an arbitrary value of alpha as a scaling factor. Blend:
o Approximation band of C and Cip,
o Horizontal detail band of C and Cip,
o Vertical detail band of and Cip,
o Diagonal detail band of C and Cip.
6. Perform Inverse DWT to combine the approximation and detail sub bands that we get after blending process to get stego image, St.
3.5.2 Decoding Process
The decoding process includes DWT, Arnold Transform, RC4, Alpha blending, IDWT. To recover the secret image we require the original cover image and the stego image. The decoding process is almost similar to encoding process in reverse direction. The schematic diagram of decoding process is shown in figure 3.3.
Figure 3.3. Decoding process of proposed Steganography
Algorithm for decoding process
1. Select stego image, St and Cover image, C.
2. Perform 2-D DWT at level-1on both C and St.
3. Perform alpha blending C and St.
4. Apply IDWT on these four sub bands. Now, what we have is the ciphered image, Cip.
5. Decrypt the Cip using RC4 decryption algorithm by inputting the key, K2. We will get scrambled image, Sc as a result.
6. Apply Anti – Arnold transform to recover the original secret image.
4. Experimental results and performance analysis
The performance of the proposed method is evaluated by calculating the parameters of PSNR, MSE, AD, MD, and SC for undetectability. In our experiment, we have tested 85 images with the proposed method. For the representation purpose, we have used house.jpg, cameraman.jpg as cover images and lena.jpg, eintein.jpg as secret images. All of these four images are of 256*256 sizes.
4.1Performance Analysis
PSNR (Peak Signal to Noise Ratio) gives the distortion between cover and stego image. Higher value of PSNR shows better quality stego image. MSE (Mean Square Error) represents the difference between corresponding pixels of original cover image and stego image [1]. The value of MSE should be lower. AD (Average Difference) shows the average difference of the pixels of cover and stego image [1]. AD should be lower in this case. MD is the maximum difference between any two corresponding pixels in the cover and stego image [1]. Its value should be lower.