Enhanced Joint and Separable Reversible Data Hiding in Encrypted Images with High Payload

Fatema-Tuz-Zohra Khanam, Sunghwan Kim
2017 Symmetry  
Recently, much attention has been paid to reversible data hiding (RDH) in encrypted images, since it preserves the data that the original image can be perfectly recovered after data extraction while protecting the confidentiality of image content. In this paper, we propose joint and separable RDH techniques using an improved embedding pattern and a new measurement function in encrypted images with a high payload. The first problem in recent joint data hiding is that the encrypted image is
more » ... pted image is divided into blocks, and the spatial correlation in the block cannot fully reflect the smoothness of a natural image. The second problem is that half embedding is used to embed data and the prediction error is exploited to calculate the smoothness, which also fails to give good performance. To solve these problems, we divide the encrypted image into four sets, instead of blocks; the actual value of pixels is considered, rather than an estimated value, and the absolute difference between neighboring pixels is used in preference to prediction error to calculate the smoothness. Therefore, it is possible to use spatial correlation of the natural image perfectly. The experimental results show that the proposed joint and separable methods offer better performance over other works. Several RDH techniques have been introduced [4-10]. In 2003, Tian [4] proposed a reversible data hiding method using the difference expansion technique, where one bit is embedded into the difference of two successive pixels. In 2006, Ni et al. [5] introduced a novel RDH system based on histogram shifting by utilizing the minimum and maximum points of the image histogram; data are concealed by shifting the histogram. In 2007, Thodi and Rodriguez [6] proposed a different technique by expanding the prediction errors. In 2010, Luo et al. [7] presented a new reversible data hiding process by adopting an additive interpolation error expansion technique, which provides very small falsification and a comparatively high capability. Moreover, to improve performance, many techniques have been proposed for typical reversible data hiding approaches [8] [9] [10] . In some applications of data hiding, the embedded carriers are further encrypted to prevent the carrier from being analyzed to expose the existence of the embedding [11, 12] . In 2007, Lian et al. [11] proposed a video encryption and watermarking scheme based on H.264, the advanced video coding (AVC) codec, where media data are encrypted and partially watermarked. In 2010, Cancellaro et al. [12] presented a commutative watermarking and ciphering scheme where a watermarked image is further encrypted to increase the security of the system. Most of the existing reversible data-hiding systems are only suitable for unencrypted covers. However, for maintaining confidentiality or protecting privacy, in many fields-like medicine, the military, and the law-a content owner may demand to encrypt the original images. Meanwhile, without knowing the encryption key and the plaintext content, a channel administrator or an inferior assistant may need to insert some additional information within the encrypted images. As an authorized receiver, it is essential that hidden information can be extracted and the original cover can be recovered error-free without any loss or distortion after image decryption and data extraction. Reversible data hiding in encrypted images (RDH-EI) gratifies these requirements. The existing RDH-EI methods can be classified into two categories: "vacating room before encryption (VRBE)" and "vacating room after encryption (VRAE)". In VRBE, the content owner creates room for embedding data in the cover image before encryption [13, 14] . As VRBE requires the content owner to do an extra preprocessing before content encryption, so this method might be impractical. In this sense, the VRAE method is more practical. In VRAE methods, the original content is encrypted by the content owner, and the data-hider embeds the additional information by modifying a small part of the encrypted data [15] [16] [17] [18] [19] [20] [21] [22] [23] . The VRAE methods can be further divided into two categories: joint method and separable method. In joint methods [15] [16] [17] [18] [19] [20] , with an encrypted image containing additional data, a receiver may first decrypt it using the encryption key, and then extract the embedded data and recover the original image from the decrypted image using the data-hiding key. In 2011, Zhang [15] proposed a novel reversible data hiding technique in encrypted images. In 2012, Hong et al. [16] improved Zhang's technique by using a side match technique. Zhang's and Hong's systems were further enhanced with other techniques [17, 18] . In these systems, the encrypted image is divided into blocks, and the spatial correlation in the block cannot fully reflect the smoothness of natural images, especially when the block size is small. In 2014, Li et al. [19] introduced a new scheme where a random diffusion strategy is used for embedding and accurate prediction is used to measure the smoothness. In 2014, Wu and Sun [20] proposed a different joint RDH system based on prediction error. However, the smoothness calculation from using prediction error fails to perform well. Nevertheless, in joint methods, the embedded data can only be extracted before image decryption. In other words, a receiver having a data-hiding key but no encryption key cannot extract any information. Moreover, when the payload is high, it is not possible to get error-free extracted bits with all these joint methods. To overcome these problems, a separable reversible data-hiding scheme is required. In separable methods [20] [21] [22] [23] , with an encrypted image containing additional data, if a receiver has the data-hiding key, he can extract the additional bits from the marked encrypted image directly, while if the receiver has the encryption key, he can decrypt the received data which is similar to the original one. When, the receiver has both the data-hiding and encryption keys, he can extract the additional
doi:10.3390/sym9040050 fatcat:rgwby6jd75buhdjri2tayklkha