2013 2nd National Conference on Information Assurance (NCIA)
Selective Region Based Images Encryption Irfan Ullah, Waseem Iqbal, Dr. Asif Masood Department of Information Security Military College of Signals (MCS) National University of Science and Technology Islamabad, Pakistan
[email protected],
[email protected],
[email protected]
After area is selected, it is encrypted using map-images. With morphological operations different regions can be selected by sender for encryption [5, 6]. In [7], selective encryption of JPEG Images uses the idea of detecting human skin in color images. The basic concept of the algorithm is to find region of interest (ROI) for further processing. While applying JPEG compression technique, two types of coefficients that are AC and DC coefficients, are proposed. AC and DC coefficients are further processed into bits with Huffman encoding. The paper is organized as follows: In subsequent session, the proposed encryption algorithm is presented. Section 3 gives some experimental results and last section concludes the paper.
Abstract— Image security is of great importance in many applications including Military, Medical and many others. Generally, images are of very large sizes and conventional encryption techniques are not considered good. An approach that is very recently found in literature is of selective encryption of multimedia images. In this paper a new approach is introduced for the selection of the area in the image that will select maximum information area for encryption based on percentage of coefficients. Un-encrypted area is permuted with the encrypted area that will further enhance the security of the images. The mean square error (MSE) and peak signal to noise ratio (PSNR) values show a huge difference in between the original and encrypted images. Keywords—AES; Permutation; Encryption
II. I.
INTRODUCTION
The proposed Selective Image encryption technique is a new method for selecting the sensitive area in the images for encryption purpose. The main idea in this approach is to selectively encrypt / decrypt the sensitive area followed by permutation. The design for proposed encryption method is shown in the Figure. 1
Usually image encryption requires a very large amount of processing that is quite demanding for an efficient solution. One way to reduce this requirement is the use of selective image encryption because there are so many applications that usually do not require the full image encryption, e.g., in the bank cheque, only the signature of the owner is important and the total amount requires security. The concept of selective encryption is used in those applications where speed and the security are the main concern. A selective/partial image encryption algorithm that uses quadtree and wavelet algorithms for compression has been proposed in [3]. The proposed encryption method only encrypts the quadtree structure while at the leaf nodes of quadtree the block intensities are not encrypted. The partial quadtree encryption method is applied to the lossless image compressions as well as the lossy compression. For transmission the orders of the leaf nodes needs to be considered. JPEG image encryption technique proposed in [1, 2, 4], selects AC coefficients after performing Discrete Cosine Transform (DCT) and Quantization for encryption and leave DC coefficients unencrypted as they are highly predictable. Two more selection encryption schemes are proposed in [5]. First method proposed map-based technique for images encryption partially whereas; second technique performs some morphological operation for the selection of sensitive area.
978-1-4799-1288-9/13/$31.00 ©2013 IEEE
PROPOSED ENCRYPTION ALGORITHM
Fig. 1. Design of proposed encryption scheme
In the proposed technique, image is divided into different blocks of size 8x8 before encryption. The idea is illustrated in Figure. 2. In this section a new approach is introduced that will randomly select the area for encryption based on the coefficients percentage of the blocks. The algorithm for determining the percentage coefficient of each block is discussed as,
125
• • • • •
Eliminate negative sign by taking absolute of the whole matrix (keep record of negative sign index) Sum up coefficients of whole matrix which is actually 100% content Sum up coefficients of individuals blocks and convert it into percentage, this will give the percentage content of each block. Find the block with the maximum percentage content. Select a threshold value as referenced from the block with maximum content.
1 2 1 7 2 5
9 3 7 9 1 6
8 6 5 9 3 7
7 9 4 8 1 1
6 3 6 8 2 3
1 2 7 6 5 4
Fig. 3. 2 X 2 Block Matrix
Percentage of block 1 = 8.62% After finding percentage content of all blocks, we obtain a block of percentages, as are show in Figure 4.
For example, maximum percentage content for some block is 1%. If threshold value is selected 20%, then block having percentage content from 0.8% - 1% will be selected for further processing and if the threshold value is selected to be 30%, then blocks having percentage content from 0.7% - 1% will be selected. After selecting the threshold all blocks falling with in the criteria are selected and their index numbers are recorded. These selected blocks will be further processed for the process of encryption using AES algorithm. After image is divided into blocks, discrete cosine transform (DCT) is applied on each block that will convert an image to frequency domain and then Quantization is applied [1], that will divide each block obtained after DCT using a quantized matrix [8, 1]. Next to quantization, encryption area is selected based on the percentage of content in each block.
8.62%
17.24%
6.89%
13.79% 14.94% 15.51% 8.04%
6.89%
8.04%
Fig. 4. Content Percentage of Blocks
All the percentage values in the block will sum up to 100%. It can be seen that block 2 is with maximum percentage content. A threshold value is selected with respect to the maximum percentage. In the proposed scheme 60% of image is selected. Threshold value is 60% of image with respect to the block with maximum percentage, Threshold value is, (60/100) * 17.24 = 10.344 The threshold value is subtracted from block with maximum percentage of content which result in 6.9%. In next step, all blocks having percentage of content ≥ 6.9% are selected and consequently encrypted with AES Algorithm. The last step of the proposed technique is the permutation. For example, a matrix of size 3 x3 is selected, which is permuted based on random matrix B. ⎡10 11 12⎤
If the original matrix is: ⎢ ⎥ and the random matrix ⎢13 14 15⎥ ⎢⎣16 17 18⎥⎦
⎡ 1 6 9⎤ is given as B = ⎢7 3 8⎥ Thus the permuted matrix will be ⎥ ⎢ ⎢⎣ 2 5 4⎥⎦ ⎡10 17 18⎤ obtained as, ⎢12 16 15⎥ ⎥ ⎢ ⎢⎣13 14 11⎥⎦
Fig. 2. Proposed Algorithm.
The content of the blocks are the quantized value after the image is converted into frequency domain and is the value of pixels in the frequency domain. These contents are sensitive as compared to other, as human eyes are sensitive to lower frequency than to high. Consider a toy example of an image with blocks shown in Figure. 3. It shows block division of an image and the values in each block are the coefficients value of each pixel after converting it from spatial to the frequency domain.
The decryption of the proposed technique is exactly the same as the encryption but all steps need to be carried out in the reverse order. The same key will be used for decryption that was used in the encryption process and if there is a single bit changed in the decryption key then it will never give the original image.
The selection criterion is to select the block with sum of its content. Here, Sum of numbers of all blocks is 174 which is the 100% content of the image Sum block 1 = 15
126
III.
EXPERIMENTAL RESULTS
was found that this algorithm select less than 50% of the blocks in the image for the process of encryption and each block contain on average of 8-9 non-zero coefficients. So the amount of data selected for encryption is very small as compared to the previous algorithms.
To demonstrate the proposed technique, experiments have been conducted on different images. The original image and its result in the encrypted form are shown in Figure. 5.
Table 1. Encryption time for full and selective images encryption Images Full Image Selective Image Encryption Time Encryption Time (Sec) (Sec)
Lena
8.5539
3.676021
Bird
14.6491
3.02769
Fig. 5. a. Input Image (MSE = 45.5038) b. Resultant Image (MSE = 8736.6388 PSNR = 8.7174
Flower
12.1516
2.83967
Skull
8.5067
2.445685
The permuted image and after permutation the decrypted image is shown in Figure. 6 along with its MSE and PSNR value. The PSNR value that is obtained in the result shows a huge difference between the original and cipher image.
Gray Ship
6.3369
2.049593
Some other experimental results are shown,
Fig. 6. (a) Original Image (b) Encrypted Image Fig. 6. a. Permuted Image (MSE= 8437.5269 PSNR = 8.8687) b. Decrypted Image (MSE = 0.00090188 PSNR = 78.5793)
The very important factor is to calculate the time taken for image encryption. Different images were selected during the experiment and the result of different experiments performed is shown in the Table 1 below. Experiments were conducted on both full images and selective images for encryption and total time taken by both full images and selective images is calculated. The results in Table. 1 shows that proposed selective image encryption takes very less amount of time in both selecting and then encrypting the image, maintaining the security of the transmitted image whereas, full image encryption takes more time as compared to the proposed selective image encryption algorithm. The results shows that proposed encryption algorithm is very efficient as compared to the full image encryption and the algorithm shows very good shuffling results ensuring the security of the images. The DC coefficient is also encrypted with the AC coefficients because it was left unencrypted due to its highly predictability after encryption but in this paper it is encrypted in the whole block. The permutation will shuffle all the coefficients. This algorithm does not select some coefficients in blocks but it will select a whole block for encryption. During experiments it
(c) Permuted Image (d) Decrypted Image
Similar experiments were conducted on other images as well including gray-scale images and results obtained showed good encryption results. The experiments were carried out on Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz 2.39GHz L2 Cache 8MB. The algorithm is secure against replacement attack and the edge detection as whole block is encrypted along with DC component and permutation is used which further shuffle the bits in the image based on random number generated. Single bit change in the encryption key shows a total different result in the decryption.
127
IV.
CONCLUSION
REFERENCES [1]
This paper proposed a selective image encryption algorithm that contains maximum of the image information. A different approach is introduced for the selection of regions that needs to be encrypted, the encryption algorithm AES is used in the encryption process. Permutation is applied to the encrypted image to further shuffle the encrypted image that will increase one more level of security to the encrypted image. Different experiments were conducted on full images and selective image encryption which shows that algorithm proposed in this paper takes less encryption time than full image encryption and key sensitivity test shows that algorithm is secure against different attacks. Experiments were conducted on JPEG images using the proposed approach for the selection of area in the image. This research work can be extended to different direction. The proposed algorithm can be used in encrypting the satellite images as satellite image information are very confidential information for military organization and need greater security when transmitted over the untrusted network during war, similarly this research can be extended to encrypt medical images because medical image information also needs greater security from transferring into unauthorized hands.
[2] [3] [4]
[5]
[6]
[7]
[8]
128
The International Telegraph and Telephone Consultative Committee “JPEG Standard (JPG ISO/IEC 10918-1 ITU-T Recommendations T 81)” at http://www.w3.org/Graphics/JPEG/itu-t81.pdf D. Taubman and M. Marcellin “JPEG2000: Image Compression Fundamentals, Standards and Practice” Page 200-246 Springer 2002. H. Cheng and X. Li, “Partial Encryption of Compresed Images and Videos,” IEEE Transactionson Signal Processing, pp. 2439-2451. M. Podesser, H.P. Shchmidt and A. Uhl, “Selective Bitplane Encryption for Secure Transmission of Image Data in Mobile Environments,” 5th Nordic Signal Processing Symposium, on board Hurtigruten, Norway, October 4-7, 2002. H.T Panduranga, S.K NaveenKumar, “Selective Image Encryption for Medical and Satellite Images”, “International Journal of Engineering and Technology (IJET), 5(1), 115 - 121” 2013. B.D Parameshachari, K.M Sunjiv Soyjaudah, M.V Chaitanyakumar, A Study on Different Techniques for Security of an Image, International Journal of Recent Technology and Engineering ™ (IJRTE), Volume- 1, Issue-6,Jan2013 J.M Rodrigues, W. Puech, A.G Bors, "Selective Encryption of Human Skin in JPEG Images", IEEE International Conference on Image Processing, 2006. J.M. Rodrigues, W. Puech, and A.G. Bors, “A Selective Encryption for Heterogenous Coloured JPG Images Basedon VLC and Advanced Encryption Standard (AES) Stream Cipher,” in CGIV’06, Leed,UK, 2006.
