Spatial domain image steganography based on right most digit replacement and parity bit differencing / Mehdi Hussain

• Main Author: Mehdi , Hussain
• Format: Thesis
• Published: 2017
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://studentsrepo.um.edu.my/8117/1/All.pdf; http://studentsrepo.um.edu.my/8117/6/mehdi.pdf; http://studentsrepo.um.edu.my/8117/

The rapid advancement in digital computation, communication and exponential proliferation of the Internet has now become the easiest and economical way of data transmission. This evolution has gained the drawbacks in the advancement of forgery tools and application that enable perpetrators to steal, alter and destroy information during transmission. However, encryption and steganography are the most effective solutions to secure sensitive data to avoid malicious and forgery activities. In encryption, sensitive information transforms into meaningless data with observable existence, while steganography embeds the secret information inside an object (i.e. image, video, text, and audio) with invisible existence. There has been noteworthy research on image based steganography techniques to overcome the various challenges i.e. lower embedding capacity, imperceptibility, embedding efficiency and robustness against steganalysis detection attacks. However, most of the existing embedding algorithms are incapable of overcoming the adverse effects of the challenges simultaneously. In general, high capacity based methods i.e. LSB substitution, employed the multi-bit-planes for concealing the secret data that eventually modify the cover pixel values and becomes more prominent although human vision are not able to identify those changes. Consequently, these pixels changes generate various detectable artifacts i.e. dissimilarity or significant difference errors between cover and stego-pixels, that eventually leads the steganalysis methods to exploit these effects to expose the presence of secret data. The proposed right most digit replacement (RMDR) method deals with these challenges by substituting the digits instead of bits. The closest selection of digits substitution reduces the differences error between cover and stego-pixels. Furthermore, the RMDR has proven to be the best alternative to the existing LSB-based substitution techniques with enhanced imperceptibility and security against regular and singular (RS) steganalysis. The second proposed method is Parity Bit adjustment in Pixel Value Difference (PBPVD). The PBPVD method efficiently exploits the Pixel Value Difference (PVD) adjustment process to conceal extra secret data with the correlation of parity bits. Consequently, it improves the embedding capacity while retaining the imperceptibility. Furthermore, the PBPVD adjustment process would be ideal to exploit in all existing PVD-based techniques to enhance its capacity without degrading the visual quality. Finally, this research proposed hybrid embedding methods by integrating the above RMDR and PBPVD techniques to achieve the optimal steganography objectives. A comparative study presented between the proposed and the existing steganographic techniques. For imperceptibility, distortion between cover and stego-images assessed by utilizing the Peak signal-to-Noise Ratio (PSNR), Mean Square Error (MSE) and Universal Quality Index analysis (Q) matrices. For security aspect, the robustness against steganalysis detection attacks evaluated by RS, pixel difference histogram, and bit-plane analysis. Results from the above evaluations have proven that proposed methods achieved the optimal performance regarding general steganography objective/challenges. In addition, the proposed methods have proven the robustness against powerful modern SPAM feature based steganalysis detection attacks at low embedding rate.