Study On Reversible Data Hiding In Encrypted Images

In the era of digital data,which has brought significant changes to our society and our lives.Digital data benefits have also introduced new problems and opportunities for innovation.The conventional way to secure digital data during passing through any insecure channel is cryptography and data hiding.Cryptography has been used to make the digital data unreadable for secure communication whereas data hiding has been used to conceal the very existence of additional data into any digital cover media(Image,Audio,Video,etc.).Cryptography and data hiding have been complementarily used to secure modern digital data communication.Popularity of data hiding techniques has been growing exponentially in the area of watermarking,steganography and other kinds of applications.Watermarking and steganography techniques leave some distortion in the cover during recovery phase.With the development of sensitive digital data in the field of medical imagery,law and forensics and content authentications etc.where distortion is inadmissible,reversibility is one feature which is required.To,support this,a new kind of data hiding technique called Reversible Data Hiding(RDH)has come into existence,where the original media and embedded data can be recovered losslessly.The prime media used in this thesis is an 8-bit greyscale image.In the present scenarios,outsourcing of digital images have become more and more popular,where the third parties are involved to hold or manage the data and provide services to access the digital data any time anywhere.Data hiding plays a significant role to embed some additional data(source information,time tagging,remark etc.)in digital images to manage it efficiently.Moreover,the content owner is concerned about the privacy of images,so the image is encrypted before sharing with third parties,which is subsequently embedded with some additional data.Thus,the need of reversibility,data hiding along with encryption of the cover leads to the technique of Reversible Data Hiding in Encrypted Images(RDHEI).This dissertation discusses theoretical and practical aspects of RDHEI and its application to multimedia security and communication for sensitive digital images.The existing RDHEI schemes face certain issues regarding low embedding rates,scalability,quality of the recovered images,inevitable data expansion.In this thesis,four schemes are presented for the purpose of resolving or improving over these issues.Inheriting the basic RDHEI structure,the schemes have three stakeholders: content owner,data hider and the receiver.Initially,the image is encrypted by the content owner and sent to the data hider where,the data hider embeds the additional data into the encrypted image.The embedded additional data and the image is recovered from the marked encrypted image by the receiver.The first proposed RDHEI scheme is separable in nature and enables embedding multiple additional data into encrypted image which can be extracted by the authorized receivers using their embedding keys(personal key).Here,the image is encrypted by the content owner using stream cipher,and the data hider embeds the multiple messages(N number of messages for N number of users)into the encrypted image with different embedding keys.The different receivers can extract their embedded information from the marked encrypted image.This scheme facilitates to manage the multiple user's information in single image on the Cloud and improve the scalability in terms of users extracting the data.The second proposed RDHEI scheme is based on symmetric key cryptography,where encryption key is used to recover the image at the receiver side.The scheme is separable in nature.Initially,the original image is preprocessed in a manner to estimate the locations suitable for embedding before encryption.Next,information of location(location map)is compressed efficiently and embedded into encrypted preprocessed image,to avoid the overhead of separately sharing it with the data hider.Furthermore,with the help of location information data hider embeds the additional data into encrypted image.At the receiver end,additional data and original image are extracted from the marked image.The scheme improved the quality of the directly decrypted image when compared to the present state-of-the-art methods.Moreover,the original image recovery is independent of the embedding key.The third and four proposed RDHEI schemes are based on asymmetric key cryptography,where public key is used for encryption and secret key also known as private key is used for decryption.In addition,asymmetric key based RDHEI schemes solved the issue of separate sharing the key with the receiver's which adds advantage over the symmetric key based RDHEI schemes.Both the schemes solved the extra data expansion on data hiding side.The third RDHEI scheme is non-separable in nature.Firstly,the original image is preprocessed in a manner to create space by converting the MSBs to zero of the appropriate pixels,where appropriate pixel is identified by the interpolation technique.Next,the preprocessed image is encrypted by the Paillier cryptosystem,which contains probabilistic and additive homomorphic property.Next,the data hider embeds the additional data into encrypted image by using additive homomorphic property of Paillier cryptosystem.At the receiver's side,the embedded additional data is extracted after decryption of marked encryption image.The scheme controls the extra data expansion in the data hiding phase.Fourth RDHEI scheme having the property of separability,solves the problem of extra data expansion at the data hiding phase.Moreover,the scheme avoids any preprocessing step at the content owner side.Each pixel of image is encrypted by the Paillier cryptosystem.The data hider uses the additive property of Paillier cryptosystem to modify the encrypted pixel losslessly in order to embed the additional data.And the embedded data is extracted directly from the marked encrypted image.Moreover,the lossless recovery of image is possible just only by using the secret key.Which means,the embedding procedure does not introduce any distortion in encrypted image.