Digital Signal Processing 21 (2011) 433–446 Contents lists available at ScienceDirect Digital Signal Processing www.elsevier.com/locate/dsp Protecting patient privacy from unauthorized release of medical images using a bio-inspired wavelet-based watermarking approach Pegah Fakhari a,∗ , Ehsan Vahedi a,b , Caro Lucas a a Electrical and Computer Engineering Department, University of Tehran, Iran b Electrical and Computer Engineering Department, University of British Columbia (UBC), Canada article info abstract Article history: Available online 21 January 2011 Keywords: Medical image Particle swarm algorithm Genetic algorithm Watermarking Wavelet transformation Patient privacy This paper identifies a novel digital watermarking approach for copyright protection and authentication of medical images based on the wavelet transformation. We consider the problem of protecting patients’ medical records and tracing illegally distributed medical images in a group communication environment. We employ the particle swarm algorithm and genetic algorithm optimization principles to obtain performance improvement in our work. In the proposed method, the strength of the embedded watermark and noise are controlled to prevent the images from being used directly regarding to visual properties of the host signal. These parameters and also the places of the embedded watermarks are varied to find the most suitable ones for images with different characteristics. The experimental results show that the proposed algorithm yields a watermark which is invisible to human eyes, robust against a wide variety of common attacks and reliable enough for tracing colluders.  2011 Elsevier Inc. All rights reserved. 1. Introduction In recent decades, widespread use of Internet and accelerated growth of development in computer applications has resulted into unlimited illegal duplication and dissemination of copyrighted ma- terials, including images, audios, and videos. Therefore authors, publishers and multimedia providers are reluctant to allow the dis- tribution of their documents in a network environment. As a result, copyright protection has become a challenging problem regarding security and illegal distribution of privately owned images [1]. Digital image watermarking declares the ownership of the orig- inal sources by embedding the secret appropriate information, or the watermark in the original image [2–5]. This must be performed in such a way that the added information does not cause serious degradation of the perceptual quality and cannot be removed eas- ily [7]. Concomitantly, transmitting medical images between hospitals and exchanging the patients’ data such as images and diagnostic reports between physicians via networks, cause complicated net- work protocol, image compression and security problems. Hospital Information System (HIS) and Picture Archiving and Communi- cation System (PACS) have been established to provide security solutions to ensure confidentiality, integrity, and authentication. Moreover, many applications of watermarking in medical images * Corresponding author. E-mail addresses: pegah.fakhari@ece.ut.ac.ir (P. Fakhari), vahedi@ece.ubc.ca (E. Vahedi), lucas@ipm.ir (C. Lucas). have been investigated extensively to assure security [8–12]. These applications consist of ‘invertible watermarks’ that can completely erased after verifying image, data hiding, integrity control and protection, authenticity and interleaving patient information with medical images. Watermarks used for these applications need to be sensitive to any modifications to the images. For purpose of trac- ing, as another feature of digital watermarking, robustness against image distortion is highly needed; so watermarks cannot be di- rectly applied [13]. In general, we can classify digital watermarking algorithms into two classes depending on the domain of watermark embedding. The first group belongs to the algorithms which use spatial do- main for data hiding [7,14–18] while algorithms of the second group take advantage of transformation domains like Discrete Co- sine Transform (DCT) [19,20], Discrete Fourier Transform (DFT) [21] and Discrete Wavelet Transform (DWT) [22–27] for watermarking purpose. Previous works reveal that transform domain schemes are typically more robust to noise, common image processing and compression when compared with spatial transform schemes [26, 28]. Digital watermarking also can be categorized into visible and invisible, fragile and robust, blind and non-blind with emphasis on authentication, rightful ownership and availability of the host im- age respectively. Based on good time-frequency features and directives that match well with the Human Visual System (HVS), using DWT is highly motivated for image watermarking among several tech- niques [26]. In medical applications, because of their diagnostic value, it is very important to maintain the quality of images. Therefore, algorithms which generally embed watermarks into co- 1051-2004/$ – see front matter  2011 Elsevier Inc. All rights reserved. doi:10.1016/j.dsp.2011.01.014