Data Hiding In Encrypted Image

Thanks to the great advances in internet technologies,mobile devices,cloud computing and social network,a large volume of data are stored online or transmitted through the network in recent years.Data owners may outsource their data to cloud storage with a lower cost and enjoy great data management service.Among the outsourced data,there is a significant volume of images which may contain lots of private information.Hence,it provokes the need of securing personal or commercial images.Encryption is one of the most common approach to solve these kinds of problem.But encrypted images are not convenient to be shared and managed data conveniently.In order to satisfy the management requirements auxiliary information such as an image's owner,storage location,timestamp,may be stayed together with encrypted images while not affecting the original contents.Basing on the above observations,the thesis will study the methods of embed data into an encrypted image.The target of data embedding in an encrypted image is to meet these requirements: security,reversibility,separability,capacity etc.Most of the known methods cannot satisfy all of the characteristics mentioned above.In order to solve these problems,a novel method is proposed in this thesis to embed data into encrypted images.Basing on reserving-roombefore-encryption framework,an original image is divided into two parts,then applying XOR and AES encrypt the two parts respectively to provide security protection.By using difference expansion and even-odd transformation operations to settle some requirements in encryption and reserving room process.The encrypted image will be used to embed auxiliary information by using a novel embedding method.This method based on LSB modification technique to carry one bit by using an encrypted pixel which is reserved for embed data.The encrypted image after embedding additional message can be reversed to extract embedded information or decrypt to restore original image without any data loss.These two schemes data extraction and image decryption are separate.In other word,with corresponding secret keys,receiver can be first extract embedded data then decrypt encrypted image or vice versa,first perform image decryption then data extraction,respectively.One special feature in procedure of decryption original image is that: the initial image can be directly restored from marked encrypted image or stego-image with only encryption key without extract embedded data(unlike the methods which have been introduced previously).This proposed RDH method also has high embedded capacity and proved by implement results.We implement the presented method with C# programming language under Windows system.Many experiments are done on two well-known image databases.The results demonstrated the effectiveness of the proposed scheme.The embedding capacity can be as high as 0.3 bpp or higher,which is larger than the related works.And the corresponding PSNR value is acceptable(PSNR = 33.09 dB with the first method and PSNR = 58.12 dB with the second method while the embedding rate ER = 0.2 bpp when applying for Lena image).At the same time,theoretic analysis is provided to illustrate the security,reversibility and separability of our method.