Development of Efficient Image Steganography System for 3D Image Models

Abstract

Secure communication between two parties present at geographically different locations has become a matter of deep concern with information and technology progressing at such a fast pace. Steganography is the science of hiding secret message inside a harmless looking file in such a way that its presence is unnoticeable by human eye. Secret message to be hidden is considered a stream of binary bits. Secret message is sent by hiding it invisibly inside an innocuous-looking cover media. When the secret message is hidden in the cover file, stego-file is obtained as a result of embedding of secret message. A typical steganography system has two phases-embedding at message-sender side and extraction at message-receiver side. Extraction is termed as blind if cover file is not required for extraction of secret bits from stego-file. A steganography algorithm is called reversible if after extraction of secret message bits, original cover file is obtained. What cryptanalysis is to cryptography, steganalysis is to steganography. Thus, steganography should be done in such a manner that it remains an onerous task for steganalysts. In 3D image steganography system, 3D mesh model is taken as the cover file and a binary stream of secret bits is hidden inside it. A 3D mesh model consists of vertices which join together to form edges. The edges enclose an area called face and these faces make surfaces. Vertices location is the secret bits hiding place in 3D image steganography. Reversible steganography having a blind extraction is a desirable feature. Also, rotation, scaling and translation of a 3D mesh model can be carried out by intruders which can destroy the hidden secret bits. Thus, it is important for a 3D steganography algorithm to safeguard the secret bits from these operations/attacks. In this thesis, a mesh traversal algorithm is proposed. It is based on BFS (Breadth First Search) algorithm. Using the proposed mesh traversal algorithm, mesh vertices are visited and referenced in the order they have been visited. Hence, the steganography system withstands vertex reordering attack. Proposed novel steganography system of data-hiding by difference shifting scheme is both reversible and has a blind extraction. In the proposed work, two connected vertices are taken for embedding secret bits. One coordinate value-pair is picked for embedding of secret bits. This choice is made using the logistic map outcome so that secret bits are hidden randomly without forming a pattern. Logistic map is used for the first time in a 3D image steganography algorithm and thus this research work will arouse readers’ interests to use one or more chaotic systems in steganography approaches. The embedding algorithm hides secret bits inside 3D mesh model by novel difference shifting approach. Average and absolute difference values of two connected vertices are obtained. Difference is modified slightly in order to hide secret bits and new coordinates value are formed using modified average and modified difference values. Adaptive embedding is done while hiding secret bits so that the distortion to 3D cover model is minimum. Secret information to be hidden inside the 3D mesh model is first encrypted before hiding. This adds an additional layer of security to the steganography system. Secret file to be embedded is image. A novel image encryption algorithm is proposed in the present work. The image encryption algorithm is based on the Lorenz-Rossler chaotic system and DNA cryptography on RGB images. Pixel values in RGB arrays and chaotic sequences are converted to DNA sequence. A total of six arrays (=3 from RGB + 3 from chaotic sequences) in DNA strands are thus obtained and performing operations on them, encrypted color image is obtained. Evaluation of proposed image encryption algorithm indicates a fairly good performance of the system. In order to safeguard 3D stego-model from rotation, scaling and translation attacks, registration of gravity centre of 3D stego-model and two vertices is done. At the receiver side, if the gravity centre differs from that at the sender side, then 3D stego-model has been attacked. The effects of rotation, scaling and translation are reversed and secret bits are extracted. A trade-off between embedding capacity and distortion to 3D stego-model is obtained in this approach. Distortions are kept to minimum without compromising on embedding capacity. Reversible 3D image steganography techniques proposed in the literature are having very low embedding capacity. This reversible data hiding approach has decent embedding capacity and also withstands RST and vertex reordering attacks. Other reversible data hiding techniques in 2D image steganography such as prediction error expansion can be modified and used for hiding secret bits in 3D mesh model. This novel steganography approach is intriguing, impelling and efficacious.