Detection of medical identity theft in medical images through digital watermarking

Abstract

The risk of medical identity theft is increased rapidly due to the advancement in technology. Accordingly, the healthcare system security and privacy becomes an important issue. This is because any alteration to medical information may cause a devastating effect to the patient. In this work, a digital watermarking that detects any attempt of altering patient data is proposed. The discrete wavelet transform of the 8 x 8 non-overlapping blocks of the medical test image is generated prior to the embedding process. A 64-bits binary equivalent of digit numbers representing patient's entry date and file ID used as the watermark is embedded inside the corresponding patient's medical image by quantizing the coefficient of the highest frequency components of each block. After the medical image is transferred to its destination, the watermark is extracted and compared to the original watermark for authentication. The average value of the peak signal to noise ratio performance metric shows that the level of imperceptibility of the technique, was found to be 82.88 dB, and also the average values of the mean square error, and that of structural similarity showed that the level of distortion of the watermarked image was found to be 7.9e-4 and 0.9112 respectively. A number of attacks which include JPEG compression, Filtration, Gaussian noise, Salt and pepper noise, and contrast enhancement were applied to the watermarked image. The proposed algorithm enables quick and excellent detection capability of any modification done to the medical image.