Authentication of Medical Images using Integer Transforms A.Kannammal 1, Dr.S.Subha Rani 2 1 Assistant professor, 2 Professor and Head, ECE Department, PSG College of Technology, Peelamedu, Coimbatore Abstract - In these days, medical exchanges between hospitals have become a vital technique as a result of the development of the latest technology in communications and computer networks. A new fragile watermarking algorithm for medical s using selective bit plane of integer wavelet transform domain is proposed and implemented. This algorithm makes it possible to resolve the security and forgery problem of the medical s. For the selection of insertion planes, n random numbers are generated, which have the integer value from 1 to 8, and the required plane is selected using the key.msb are given as input to obtain the hash value. The output of the hash function is embedded into the selected plane, and it is combined with the MSBs to get the watermarked coefficients. The experimental results show a very good improvement in quality in terms of PSNR compared to Reversible watermarking technique using Hash algorithm. This Algorithm shows an excellent Tamper detection capability and shows the exact location of the tamper. In addition, by using integer wavelet transform, this algorithm can utilize Hash function and improves the security of the inserted watermark. Keywords - Discrete wavelet Transforms, DICOM, fragile watermark, PSNR I. INTRODUCTION Biomedical security is an important field of research in medical informatics, which is continuously growing. The increasing adoption of information systems in healthcare has led to a scenario where patient information security is more and more being regarded as a critical issue. Allowing patient information to be in jeopardy may lead to irreparable damage, physically, morally, and socially to the patient, potentially shaking the credibility of the healthcare institution. Medical s play a crucial role in such context, given their importance in diagnosis, treatment, and research. Therefore, it is vital to take measures in order to prevent tampering and determine their provenance. This demands adoption of security mechanisms to assure information integrity and authenticity. However, the security problems have emerged seriously in network transferred medical Tampering medical may cause serious results in medical treatments. With this reason, preventing medical from forgery becomes very important, and a digital watermarking technique can be a solution in resolving such a problem. This shows that the demand for medical authentication have become enormous. II. METHODOLOGY A new fragile watermarking algorithm for medical s is proposed. This algorithm makes it possible to resolve the security and forgery problem of the medical s. Instead of the discrete wavelet transform, an integer wavelet transform is used to utilize hash function. The watermark associated with the hash values is inserted into the LSBs of the integer wavelet transform coefficients. This algorithm also detects a forged area of the very well. The conventional fragile watermarking methods have handled a watermark for ; Most of them insert the watermark into the LSBs of cover. In this case, they have a problem that the watermark can be removed easily by modifying LSBs. To overcome such a problem, a new fragile watermarking algorithm using selective bit plane mechanism is proposed.the proposed algorithm uses bit plane schema in integer wavelet transform domain and solves the LSB inserting problem, which is used in conventional fragile watermarking techniques. In this technique, the inserted bit plane is selected randomly with a key. It also utilizes integer wavelet transform domain instead of spatial one. One advantage of using an integer wavelet transform is that the transformed coefficients are integers, which can be used in a hash function. An integer wavelet transform is used in this algorithm instead of discrete wavelet transform because the Hash function can encode integer values only. The coefficients of discrete wavelet transform are not integers, and the coefficients have to be changed into integers in order to utilize hash function. Making the coefficients integers by converting the real values may cause error when the inverse transform is performed. On the other way, the coefficients of integer discrete wavelet transform are integers, and these coefficients can be used in hash function encoding directly without any problems. This also makes it easier in getting a digital signature. 10
2.1. Fragile Watermarking It is important to detect and localize even the slight changes of an for authentication. In this case, the watermarks must be destroyed easily by most of attacks. These kinds of watermarks are known as fragile watermarks. The fragile watermark should have the following properties: (1) Detection of tampering: It is the most fundamental characteristic of a fragile watermark. It should detect any modification or forgery in a watermarked. (2) Perceptual transparency: It should not visible under normal observation. () No requirement of the original : It has to extract the watermark from the watermarked without the original one. () Detection of location: It means the capability of locate the changed regions within the watermarked. 2.2. Watermark insertion 1. The medical is transformed into the timefrequency domain using IWT 2. Low -low (LL) band is chosen after the first-level IWT of an input cover.. For the selection of insertion planes, n random numbers are generated, which have the integer value from 0 to k, and select plane using the key.. Input MSB values into a hash function and get hash value. 5. The output of the hash function is embedded into the selected plane, and it is combined with the MSBs to get the watermarked coefficients. Inverse integer wavelet transform is done to get the watermarked. Input MSB values into a hash function and get hash value.. Compare extracted LSBs with hash value. 5. If the integer wavelet coefficients are changed by any attacks, the hash value will be changed. 6. This change makes it possible to detect any changes of the corresponding coefficients and the corresponding locations in the spatial domain. LL Band Select Plane Key Hash Function Plane Extract MSB Extract LSB Compare Fig 2. IWT - Watermark extraction process III. RESULTS AND DISCUSSIONS Extracted The test s used are 256X256 sized JPEG and DICOM s. The test samples used are shown in fig.medical authentication using integer wavelet transform and hash function and the results are shown below LL Band Select Plane Hash Function Plane Extract MSB Embedding Key Fig.1. IWT - Watermark embedding process 2.. Watermark extraction 1. The watermarked medical is integer wavelet transformed as did in the insertion procedures. 2. Select bit plane of each block using the key. Fig. Results for 16-Ankle (DICOM) 105
7 IWT 6 5 2 1 0 PSNR/10 Time elapsed Fig 5 Results for -MRI (JPEG) Fig 7. Comparison graph for IWT.1 Discussion for IWT algorithm The test s used in this algorithm were three 256X256 sized JPEG s and three 256X256 sized DICOM s. This algorithm is invertible and it supports all data formats such as JPEG, DICOM, TIFF, GIF, etc. The simulation platform used was MATLAB R2007b. This algorithm also has excellent tamper detection capability and it shows the exact location of the tamper. As the watermark is embedded in a selective bit plane of the MSBs, this algorithm improves the security of the watermark to a greater extent. The simulation results are shown in table 2. The PSNR values are equally good for both JPEG and DICOM s. The time elapsed is also very low for all the s. This can be observed clearly from the graph shown in fig7..2. Comparison of performance metrics of RWA, ICA and IWT based algorithms Fig 6.Results for (JPEG) Para meter s CT- Table 1 Comparison of performance metrics for IWT -MRI Tran svers e brain Ankl e CT- Brai n Brai n- CS Lung s 0.07 0.060 0.0805 0.057 0.07 0.061 0.062 PSN R (db) Time elaps ed (s) 59.09 60.068 59.07 60.77 59.0 9 60.26 60.190.176.615.562.156.176.260.65 10 8 6 2 0 Lungs RWA IWT ICA PSNR/10 Time elapsed RWA- Reversible watermarking using Hash algorithm. IWT-Integer wavelet and Hash algorithm. ICA-Independent component analysis. Fig. 8. Comparison graph of three algorithms for LUNGS 106
Sl.N o 1 Wavele t used Discrete (RWA) International Journal of Emerging Technology and Advanced Engineering Table2 Overall comparisons of performance metrics Test samples Lungs 2 MRI 5 Type JPEG JPEG PSNR (db) Time elaps ed(s).16.129.211.071 2.055.09 JPEG.5656 2.6095. DICOM 0.0021 2.896.91 DICOM 0.00 0.5.702 6 CS DICOM 0.002 2.56.975 7 Discrete (ICA) Lungs JPEG 0.010 67.001 9.885 8 MRI JPEG 0.0015 76.9 10.077 9 10 11 JPEG 0.016 66.0099 10.05 DICOM 0.08 57.987.58 DICOM 0.085 56.872.1 12 CS DICOM 0.085 52.981.2 1 Integer Lungs JPEG 0.0622 60.1901.65 1 MRI JPEG 0.060 60.068.615 15 16 17 JPEG 0.0805 59.072.562 DICOM 0.057 60.771.156 DICOM 0.075 59.092.176 18 CS DICOM 0.0612 60.262.260 On comparing the performance metrics of the three different algorithms, it is found that PSNR values for ICA and IWT algorithms are far better when compared to the Reversible watermarking algorithm.i.e., they show better improvements in quality in terms of PSNR with an increase of about 15 to 20dB than the Reversible watermarking algorithm. In terms of computation time, the ICA based technique takes a longer time. This is mainly due to the algorithms used to compute independent components. This can be inferred by comparing the time elapsed with the other algorithms. The overall simulation results of the three algorithms were tabulated in table 6.. In Reversible watermarking using Hash algorithm, when DICOM s are used, watermarked s of very poor quality were obtained. This algorithm did not support all the data formats. The average PSNR of this algorithm comes around 2 db. In ICA algorithm, the time taken for computation is very large. When DICOM s are read, watermarked s showed a very poor quality. The average PSNR of this algorithm comes around 69 db. In IWT based algorithm, the average PSNR is close to 60 db. It effectively processes the DICOM s. Yet another important advantage of IWT is that, when an is transformed into a wavelet domain using DWT, the values of the wavelet coefficients will be the floating point. If these coefficients are changed during the watermark embedding, the corresponding watermarked block will not have accurate values. Any truncation of the floating point values of the pixels may result in loss of information and may ultimately lead to the failure of the reversible authentication watermarking systems, that is, the original cannot be exactly reconstructed from the watermarked. Information may be lost through forward and inverse transforms. Furthermore, conventional wavelet transform is in practice implemented as a floating point transform followed by a truncation or rounding since it is impossible to represent transform coefficients in their full accuracy. Hence information is potentially lost through forward and inverse transforms. To avoid this problem, an invertible integer to integer wavelet transform based on lifting is used here. It maps integers to integers and does not cause any loss of information through forward and inverse transforms. 107
IV. CONCLUSION Three different algorithms(reversible watermarking using Hash algorithm, Content based watermarking using independent component analysis, Medical authentication using Integer wavelet transform and Hash algorithm) have been analysed for authentication of medical s. In medical imaging, picture archiving and communication systems (PACS) are computers, commonly servers, dedicated to the storage, retrieval, distribution and presentation of s. The medical s are stored in an independent format. The most common format used in hospitals for storage is DICOM (Digital Imaging and Communications in Medicine).From the comparative studies, it has been found that Integer wavelet transform and Hash algorithm is most suitable for the authentication of DICOM s. The PSNR obtained using this algorithm is close to 60 db and shows better improvements in quality. By using integer wavelet transform, this algorithm can utilize Hash function and it improves the security of the inserted watermark. Both ICA and IWT shows excellent tamper detection by locating the exact position of the tamper. In terms of computation time, the ICA based technique takes a longer time, whereas IWT based algorithm has a little elapsed time. REFERENCES [1] Diljith M. Thodi and Jeffrey J. Rodriguez, "Prediction-Error Based Reversible Watermarking," Proc. 200 IEEE International onference on Image Processing, Oct. 2-27, 200, Singapore, vol., pp. 159-52. [2] R. C. Calderbank, I. Daubechies, W. Sweldens, and B.Yeo, Transforms that Map Integers to Integers, Applied and Computational Harmonic Analysis, vol. 5, pp. 2-69, 1996 [] Cheng-Ri Piao, Dong-Min Woo, Dong-Chul Park, and Seung-Soo Han, Medical Image Authentication Using Hash Function and Integer Transform, Congress on Image and Signal Processing, 2008. [] Aapo Hyvarinen, Survey on Independent Component Analysis, Neural Computing Surveys, vol. 2, pp. 9-128, 1999. [5] Hyvarinen, Karhunen, and Oja, Introduction, Chapter 1 in Independent Component Analysis, John Wiley, pp. 1-12, 2001. [6] Francisco J. Gonzalez-Serrano, Harold. Y. Molina-Bulla, and Juan J. Murillo- Fuentes, Independent component analysis applied to digital watermarking, International Conference on Acoustic, Speech and Signal Processing (ICASSP), vol., pp. 1997-2000, May 2001. [7] José Alberto Martínez Villanueva, Claudia Feregrino Uribe and Jezabel Z. Guzmán Zavaleta, Watermarking algorithms analysis on radiological s, International Conference on Electrical Engineering, Computing Science and Automatic Control, pp. 298-0, 2008. [8] A. Giakoumaki, S. Pavlopoulos, D. Koutsouris Secure and efficient health data management through multiple watermarking on medical s Med Bio Eng Comput pp. 619 61, 2006. [9] C.S. Woo, J. Du, and B. Pham Multiple Watermark Method for Privacy Control and Tamper Detection in Medical Images WDIC2005 pages pp. 59-6, Australia,February, 2005. [10] A. Giakoumaki, S. Pavlopoulos, D. Koutsouris A medical watermarking scheme based on wavelet transform IEEE EMBS Annual International Conference,Cancun, Mexico pp. 17-21, September, 200 [11] A. Giakoumaki, S. Pavlopoulos, D. Koutsouris, 2006, Secure efficient health data management through multiple watermarking on medical s, Med. Biol. Eng. Comput., 619 61. [12] Wei-Hung Lin, Shi-Jinn Horng, Tzong-Wann Kao, Pingzhi Fan, Cheng-Ling Lee, and Yi Pan,, 2008 An Efficient Watermarking Method Based on Significant Difference of Coefficient Quantization, IEEE Trans. on Multimedia, vol. 10, no. 5, pp. 76-757. [1] Byung S. Kim, Sun K. Yoo, and Moon H. Lee, 2006, - Based Low-Delay ECG Compression Algo rithm for Continuous ECG Transmission, IEEE Trans. on Information Technology in Bio-Medicine, vol. 10, No. 1, pp. 77-8 [1] Mustafa Ulutas, Guzin Ulutas, Vasif V. Nabiyev, Medical Image Security and EPR Hiding Using Shamir s secret Sharing Scheme, The Journal of Systems and Software 8, pp. 1-5, Dec 2010. [15] Dalel Bouslimi, Gouenou Coatrieux, Christian Roux, A joint encryption/watermarking algorithm for verifying the reliability of medical s: Application to echographic s, computer methods and programs inbiomedicine 106, pp. 7-5, 2012 108