- Open Access
- Total Downloads : 18
- Authors : Kusuma Prabhu, Raksha
- Paper ID : IJERTCONV2IS13101
- Volume & Issue : NCRTS – 2014 (Volume 2 – Issue 13)
- Published (First Online): 30-07-2018
- ISSN (Online) : 2278-0181
- Publisher Name : IJERT
- License: This work is licensed under a Creative Commons Attribution 4.0 International License
Efficient Medical Image watermarking with Tamper Detection and Recovery
M.Tech student, Dept. of E&C, NMAMIT, Nitte, Udupi, India firstname.lastname@example.org
ech student, Dept of E&C, NMAMIT, Nitte, Udupi, India email@example.com
AbstractA block-wise and content-based medical image authentication scheme with location and recovery is presented. This paper presents authenticity and integrity of medical images using watermarking. In this section a case of using intensity average comparisons and parity bits as the authentication watermark is presented. To localize tamper in a block, the watermark needs to be embedded directly into that block. If a block is being tampered locally, the intensities of the pixels involved will be changed. This will also change the average intensity of the block concerned. To ensure that this is not changed, a parity check will be used. However, a parity check alone will not guarantee that the block has not been changed. To overcome this, the intensity comparison is used as another guard. Watermarking should also serve as an integrity control and should be able to authenticate the medical image Authentication Watermarking with Tamper Detection and Recovery is able to localize tampering, while simultaneously reconstructing the original image.
Keywordsmedical imaging, integrity control, watermarking
Medical Images such as radiographs, ultra sound and magnetic resonance images play important part in the process of diagnosing a patient by medical practitioners. Advancement in the medical information system is changing in the way patient records are stored, accessed and distributed. Medical images can be stored in the digital form temporarily or permanently on a server along with patient records. Malicious tampering of medical image for the purpose of Insurance claims or to hide a medical condition for personal gain is possible. The integrity of the medical images and their information needs to be protected from unauthorized modification or destruction. The current security measures used to protect integrity of the patient records are such as VPN (Virtual Private Network, Data encryption, Data embedding. It is vital to keep images safe from any damage, it is also important being able to detect when an image has been modified. Indeed, medical images can be modified accidentally, for example during their transmission, or deliberately. In this latter case, images can be tampered with the introduction or the removal of lesions. Also, it must be known that some image processing may lead to similar situations. In telemedicine applications, for instance, lossy image compression is tolerated so as to reduce the amount of information to be transmitted. However, depending on its extent, this process may induce unacceptable information loss and results in a misdiagnosis, involving at the same time liabilities of physicians.
In this section, an efficient and effective digital watermarking method for image tamper detection and recovery is presented. The method is based on four concepts introduced from the literature: 1) block-based Fridrich et al. 2) Separating authentication bits and recovery bits Lin et al. 3) Hierarchical Celik et al  and Average intensity as an image feature Lou et al. The method is efficient as it only uses simple operations such as parity check and comparison between average intensities. It is effective because the scheme inspects the image hierarchically with the inspection view increasing along with the hierarchy so that the accuracy of tamper localization can be ensured. This scheme can perform both tamper detection and recovery for tampered images. Tamper detection is achieved through a block-based, inspection and recovery of a tampered block and relies on its feature information hidden in another block, which can be determined by a one-dimensional transformation.
THE PROPOSED AUTHENTICATION METHOD LSB is suggested, to minimize the degradation as
medical images are very strict with the quality. The recovered image, however, will not be considered authentic and will not be used for any clinical purposes. One possibility for the purpose of recovery is to help in the investigation to find the motive and the person responsible for the tampering. A 3×3 sub block in a 6×6 block is suggested to accommodate two authentication bits and seven recovery bits to be embedded in the LSB of each pixel.
For each block B of 6×6 pixels, divide it into four sub- blocks of 3×3 pixels. The watermark in each sub-block is a 3-tuple (v, p, r), where both v and p are 1-bit authentication watermark, and r is a 7-bit recovery watermark for the corresponding sub-block within block A mapped to B. The following algorithm describes how the 3-tuple watermark of each sub-block is generated and embedded which is described in fig 1 and fig 2.
Set the LSB of each pixel within the block to zero and compute the average intensity of the block and each of its four sub-blocks, denoted by avg_B and avg_Bs, respectively.
Generate the authentication watermark, v, of each sub- block as:
1 if avg_Bs avg_B
0 otherwise (1)
Generate the parity check bit, p, of each sub-block
1 if number is odd,
Where, num is the total number of 1s in the seven MSBs of avg_Bs.
From the mapping sequence generated in the preparation step, obtain block A whose recovery information will be stored in block B.
Compute the average intensity of each corresponding sub-block As within A, and denote it avg_As.
Obtain the recovery intensity, r, of As by taking the seven MSBs in avg_As.
Embed the 3-tuple watermark (v, p, r), 9 bits in all, onto the LSB of each pixel in Bs.
Fig 1 Watermark generation and embedding location
Fig 2 AW-TDR embedding scheme
The test image is first divided into non-overlapping blocks of 6×6 pixels, as in the watermarking embedding process. For each block denoted as Br, the LSBs of each pixel in Br were set to zero and compute its average intensity, denoted as avg_Br. A 2-level detection is then performed. In level-1 detection, each 3×3 sub-block within one block is examined. In level-2 detection, a 6×6 block is treated as one unit. Level-3 detection is for VQ attack resilience only. The procedure of our hierarchical tamper detection scheme is described in the following:
For each sub-block Brs of 3×3 pixels within the block Br, perform the following steps:
Extract v and p from Brs.
Set the LSBs of each pixel within each Brs to zero and compute the average intensity for each sub-block Brs, denoted as avg_Brs.
Count the total number of 1s in avg_Brs and denote it as Ps.
Set the parity check bit p of Brs to 1 if Ps is odd, otherwise, set it to 0.
Compare p with p. If they are not equal, mark Brs as tampered and complete the detection for Brs.
Set the algebraic relation v=1if avg_Brs>=avg_Br, otherwise, set it to 0. 7. Compare v with v. If they are not equal, mark Brs as tampered and complete the detection for Brs; otherwise mark it as valid.
For each block of size 6×6 pixels, mark this block tampered if any of its sub-blocks is marked tampered; otherwise mark it valid.
For each valid block Br f size 6×6 pixels, perform the following steps:
Find the block number of block C, where block C is the one in which the intensity feature of block Br is embedded.
Locate block C.
If block C is marked tampered, assume block Br is valid and complete the test.
If block C is valid, perform the following steps: Obtain the 7-bit should-be intensity of each Brs by extracting the LSBs from each pixels in the corresponding block within block C, padding one zero to the end to make an 8-bit value.
Compare with avg_Brs and mark Br tampered if they are different.
After the detection stage, all the blocks are marked either valid or tampered. Only the tampered blocks are recovered and the valid blocks are left as they are. For convenience, we call the tampered block, block B and the block embedded with its intensity, block C. The restoration procedure for each tampered block is described as follows:
Calculate the block number for block C.
Locate block C.
Obtain the 7-bit intensity of each sub-block within block B, padding one zero to the end to make an 8-bit value.
Replace the new intensity of each pixel within the sub-block with this new 8-bit intensity.
Repeat step 3 and 4 for all sub-blocks within block B.
The localization accuracy and recovery correctness were tested by making various modifications to the watermarked image.
Different image modularitys have been considered:
-Magnetic resonance (MRI) of head 256Ã— 256 pixels
-X-Ray imaging: Mammograms of 256Ã— 256 pixels
-Ultra sound imaging: Echo of vein 256Ã— 256 pixels
Fig. 3 Samples of our experimental image data sets
Fig 4.a. Watermarked image b. Tampered image
c. Recovered image
In Fig 4.b.The cyst was removed from the image by using the healing brush tool. If this image were a critical piece of evidence in a legal case or police investigation, this form of tampering might pose a serious problem.
Fig 5. a. Watermarked image b. Spread tampered image
c. Recovered image
Fig 6. a. Watermark image b. Salt &Pepper attack
c. Recovered image
When, Salt & Pepper noise is added in the medical image at 3% as in Fig 5.b and when image undergoes spread tampering as in Fig 6.b, the results show that our proposed algorithm has strong robustness against noise and spread tampering attacks.
The PSNR and MSE of the recovered image
In this paper we proposed a watermarking scheme that can detect and localize tamper and recover the images. The purpose is to verify the integrity and authenticity of images. We presented our watermarking procedures that include data embedding, tamper detection and recovery procedure. The experimental results demonstrate that the precision of tamper detection and localization is close to 100% after level-2 detection. The tamper recovery rate is better than 86% for a less than half a tampered image.
Fridrich, J. and Goljan, M., 1999. Images with self-correcting capabilities, IEEE International Conference on Image Processing, 3, pp. 792-796.
Lin, C and Chang, S 2001.A robust image authentication method distinguishing JPEG compression from malicious manipulation, IEEE Transactions on Circuits and Systems for Video Technology, 11(2), pp. 153-168.
Celik, M.U., Sharma, G., Tekalp, A.M., 2002. Hierarchical watermarking for secure image authentication with localization, IEEE Transactions on Image Processing, 11(6), pp.585-594.
Lou, D. C. and Liu, J. L., 2000. Fault resilient and compression tolerant digital signature for image authentication. IEEE Transactions on Consumer Electronics, 46(1), pp. 31-39.
H.-J. He, J.-S. Zhang, and H.-M. Tai, Self-recovery Fragile Watermarking Using Block-Neighborhood Tampering Characterization, in Proc. Information Hiding, 2009, pp. 132-145.
Phen-Lan Lin, Po-Whei Huang, An-Wei Peng, A Fragile Watermarking Scheme for Image Authentication with Localization and Recover, Multimedia Software Engineering, 2004. Proceedings. IEEE Sixth International Symposium on 13-15 Dec. 2004.