Reversible Data Hiding In Encrypted Domain

Reversible data hiding in encrypted domain(RDH-ED)is an important branch of the reversible data hiding field.It can embed external data into the encrypted image,which not only protects the user's personal privacy,but also achieves the purpose of content marking or content authentication.Currently,there are two types of RDH-ED:(1)Non-separable RDH-ED,(2)separable RDH-ED.The second type is the main research content of RDH-ED,which is also the most widely used.However,the existing separable RDH-ED algorithms suffer from problems such as low embedding rate and poor security.Furthermore,existing separable algorithms mainly use two-dimensional(2D)images as carriers,and there are few studies on three-dimensional(3D)volume data.However,3D volume data has been widely used in the medical image processing,seismic exploration and virtual reality fields.This thesis has done the following works on the above issues:(1)Divide the image into blocks.The pixels in each block are divided into two groups,where the pixels in the first group are called sample pixels and the pixels in the second group are called non-sample pixels.The sample pixels are used to predict non-sample pixels.In the image encryption process,the image is encrypted using a combination of stream cipher and block scrambling.It masks the contour of the original image and increases the security of the secret image.(2)A high embedding rate scheme based on bit replacement is proposed.External data is embedded in redundant bits of the prediction error.The prediction errors of the non-sample pixel position are preprocessed.The highest bit of the binary form is used as the positive and negative identification to mark the prediction error.The most significant bit in binary form is used as a positive and negative flag to mark the prediction error.The processed prediction errors are placed at the corresponding non-sample pixel positions.Since most prediction errors can be represented using the highest bit and the lowest four bits of the binary form,the remaining bits can be used to embed external data.This solution has the advantages of built-in positioning map,high embedding capacity and easy extraction.(3)According to the characteristics of 3D volume data,three prediction schemes for 3D volume data are designed and compared,which are 2D prediction,3D prediction and layer prediction.According to different application requirements of the recipient,the embedding scheme based on bit replacement and the embedding scheme based on prediction error expansion are applied to encrypted 3D volume data.A large amount of external data can be embedded using the embedding scheme based on bit replacement.It has a high embedding rate.Although the embedding rate based on prediction error expansion is low,the external data can be extracted on the decrypted image.And it extends the application scenario.The experimental results show that the image encrypted by this scheme has lower similarity with the original image.It has better masking effect on image contours and has higher security performance than previous algorithms.Embedding external data with a bit replacement scheme greatly increases the embedding capacity.