Secured reversible color image data hiding technique using image classifiers and Lempel-Ziv-welch image compression technique

摘要

Recent advancement in data transfer and networking techniques has put forward a considerable threat for secure data transfer. It is the sensitive information that flows via network fuels the engine of global economy. One of the main concerns in data communication is the ability to exchange information in a secured fashion and embed the information of interest in any multimedia carrier like audio, video and an image. The proposed work is an ideal modernistic novel approach for secured sensitive information communication over an encrypted color host image carrying exceptionally confidential data. Distortion less retrieval of both payload and host signal information from marked image is an appealing feature in scenarios like medical, Military and satellite applications. Reversibility not only assures zero error retrieval of sensitive information hidden and also perfect reconstruction of host medium information contents while safeguarding the confidentiality of secret information. Most popular and widely in use Advanced Encryption Standard(AES) stream cipher in Counter mode is used for encrypting the cover image content, by performing XOR operation over cover image information bits with key dependent pseudorandom bits. Signal Processing over the encrypted domain is one of the most demanding features for most of the privacy preserving applications like cloud computing and remote sensing. High Embedding capability is achieved through Lempel-Ziv-Welch (LZW) compression technique. High performance reversible data hiding technique is assured via public key modulation scheme. Two of the most powerful image classifiers Support Vector Machine (SVM) and K- Nearest neighbor (KNN) algorithms are used at the decoder end to distinguish between encrypted and non encrypted image blocks. Performance evaluation of image classifiers is done, considering their ability to accurately categorize image patches as encrypted and unencrypted using feature vectors. Features used for categorizing encrypted and unencrypted image blocks are variation of pixel intensity in all four directions, entropy, standard deviation and histogram plot of segmented image blocks. Proposed algorithm comes with a unique feature of simultaneous retrieval of both host image and payload information in an error free fashion with zero distortion. Proposed algorithm is proven more secured considering several security attacks as evaluation parameters. Few of Cryptanalysis and Steganalysis techniques considered to verify the security feature of proposed algorithm are Sample pair analysis (SPA), Number of changing pixel rate (NPCR), Unified averaged changed intensity (UACI) and Chi-square attack.