2018 Third Scientific Conference of Electrical Engineering (SCEE), University of Technology - Iraq 978-1-7281-1587-0/18/$31.00 ©2018 IEEE 105 Robust Image Watermarking Against Crop and Rotation Attacks for Public Communication Networks Ahmed Toman Thahab Department of Electric and Electronic College of Engineering, University of Kerbala Kerbala, Iraq toeahmed@gmail.com Haider Ismael Shahadi Department of Electric and Electronic College of Engineering, University of Kerbala Kerbala, Iraq Haider_almayaly@yahoo.com Abstract— In public communication networks such as the internet, users can easily share and access multimedia products such as images, videos, and audios. Copyright protection is a vital issue to prove the ownership of products in these networks. Consequently, digital watermarking that hide invisible watermark or logo plays a significant role in overcoming copyright issue. However, any person has access to multimedia products can simply use these products. Also, there are professional attackers can remove the watermark or logo without loss the visibility of the multimedia product and claim their ownership for the digital products. This study proposes an image watermarking model that only allows authorized users to use the watermarked image. This model is robust against crop and rotation attacks. It is based on the proposed embedding method that we named a “concentric rectangles”, which embeds data in rectangles starting from the center toward outer boundary to overcome the crop and rotation attacks. The resulted watermarked image is distorted in such a way to be recognized by any user, but only the authorized one can obtain a high quality of the watermarked image. The experimental results of the proposed model show that the embedding capacity can reach up to 25% from the size of the host image with watermarked image quality above 40 dB in terms of peak signal to noise rotation (PSNR). Furthermore, the proposed model is fully retrieval for the watermark image (logo) and the logo still retrieved a with a recognized vision until 10 dB of signal to noise ratio of noise adding. Keywords— image watermarking; concentric rectangles; crop attack; rotation attack; robustness; embedding capacity. I. INTRODUCTION Communication technology is rapidly developing between users around the globe which had provided enormous accessibility to share multimedia information such as video, audio on public networks [1]. This perspective created complications since this information can be copied or altered easily without any loss of quality. It is essential to prevent unauthorized distribution of data or misuse over the internet, therefore; a revolving field of data protection known as “Watermarking” has become under the scope of research for many years [2]. Digital watermarking is a technique that embeds secret data in a digital media which can be extracted in a later stage for copyright proof; the secret data ought to be detectable even if the digital media had been manipulated under attacks [3]. This perspective implies that watermarking techniques ought to be robust against any image processing attacks such as cropping, compression,and rotation. Eavesdroppers intend to attack the watermarked media using these kinds of attacks to sabotage the mark embedded in media. This results in an intellectual property loss for authors and loss of copyright. Watermark embedding can either be embedded in the spatial domain or the transform domain. Spatial domain embedding provided less computation but degraded robustness [4,5]. In order to protect confidential information, other security techniques are combined (such as visual cryptography, ciphering…etc.) with watermarking schemes [6]. [7] Proposed a watermarking visual cryptography scheme based on repetition. A watermark is repeated in the image and adds portions of the watermark into the edge blocks of the image in order to increase robustness against cropping attack. The host image in this proposed scheme requires a pre-process before concealing the watermark. Authors in [8] proposed a watermarking scheme based on visual cryptography technique and torus automorphism. Although the authors overcome the drawbacks in [7]] of thepre-process for the host image, robustness of the technique is degraded in high compression ratios in addition to host image size restrictions. On the other hand, various schemes operate in the frequency domain to conceal the watermark using a discrete cosine transform (DCT) or Discrete wavelet transform (DWT) [9]. These schemes are considered more robust against attacks since the secret mark is scattered across the image. Wavelet- based watermark is a popular approach due to its capability to provide superior time-frequency localization [10,11]. The transform approach implies transforming the original image into a set of frequency domain coefficients and a watermark is embedded in the frequency bands. [12] Proposed a wavelet- based watermarking scheme to embed a watermark in the host image. A secret key for ownership verification is generated from the intrinsic features of the host image. The proposed scheme is vulnerable to geometric attacks such as cropping and rotational attacks. [13] Proposed a watermarking scheme using local features extracted from components of the host image in the wavelet domain. The scheme depends on a secret key derived during embedding. Work presented in [13] is capable