۱- Introduction

۲- Proposed scheme

۳- Experimental results and analysis

۴- Conclusions

References

Abstract

In this paper, an adaptive reversible data hiding scheme for encrypted images is proposed. Content owner uses an analogues stream-cipher and block permutation to encrypt non-overlapping blocks of original image through encryption key. Then, data hider classifies encrypted blocks into two sets corresponding to smooth and complex regions in original image. With data-hiding key, spare space is vacated to accommodate additional bits by compressing LSBs of the block set corresponding to smooth region. Separable operations of data extraction, direct decryption and image recovery are conducted by receiver according to the availability of encryption key and data-hiding key. Through an accurate prediction strategy, perfect image recovery is achieved. Since only a portion of blocks are modified during embedding, the directly-decrypted image quality is satisfactory. Also, more bits can be embedded into the blocks belonging to smooth set, hence, embedding rate is acceptable. Experimental results demonstrate the effectiveness of our scheme.

Introduction

Information hiding technique, also called as data hiding, has been widely studied in both academia and industry in recent years, which can embed additional data into cover data, including text, audio, image, and video, in an imperceptible way [1, 2]. There are two main research directions for data hiding: 1) achieving various protecting functionalities (copyright identification, tampering recovery, retrieval and etc) for cover data through embedding the data, i.e., watermark, in different manners; 2) realizing covert communication (steganography) for large hiding capacity of secret additional data while maintaining acceptable fidelity of cover data with well-designed encoding strategies [3-5]. The data embedding process inevitably introduces distortions on the cover image, therefore, many investigations have been carried out to study the problem of complete recovery for cover image after embedded data are extracted, which is known as reversible data hiding (RDH) [6, 7]. The embedding strategies of reported schemes for RDH can be categorized into three main types: lossless compression [8], difference expansion (DE) [9-11] and histogram shifting (HS) [12-15]. With the aim of enhancing embedding rate and stego-image quality, many studies have investigated introducing the prediction strategy into RDH [11, 13-15]. Rather than directly applying the original image as the cover data, the relative data of the original image, i.e., prediction error (PE), was constructed as the cover data for hiding.