Open AccessDEVELOPMENT OF A STEGANOGRAPHIC METHOD FOR DIGITAL IMAGES BASED ON FOURIER TRANSFORM
Author(s) -
A.V. Akhametieva,
M.D. Bezsonova
Publication year - 2020
Publication title -
zbìrnik naukovih pracʹ vìjsʹkovogo ìnstitutu kiïvsʹkogo nacìonalʹnogo unìversitetu ìmenì tarasa ševčenka
Language(s) - English
Resource type - Journals
eISSN - 2519-481X
pISSN - 2524-0056
DOI - 10.17721/2519-481x/2020/69-07
Subject(s) - steganography , computer science , block (permutation group theory) , fourier transform , digital watermarking , algorithm , channel (broadcasting) , discrete cosine transform , information hiding , binary number , embedding , artificial intelligence , image (mathematics) , mathematics , arithmetic , telecommunications , geometry , mathematical analysis
In the modern Internet space in conditions of continuous exchange of information the need to protect it from the possibility of unauthorized use by third parties is increasing. This can be ensured by using steganographic methods that allow organizing a covert communication channel.In the article a new steganographic method for digital images based on the fast Fourier transform developed. As additional information you can use both text messages and images translated into a binary sequence. For embedding of additional information 2×2 blocks are used for which the fast Fourier transform is calculated. Message embedding is done by modifying the absolute difference between two coefficients of block’s Fourier transform. Up to four bits of a binary sequence can be embedded in each block.The results of computational experiments aimed at assessing the effectiveness of the proposed method have shown that high quality of stegos is provided (average PSNR values are 58-60 dB) compared to modern analogues while ensuring high capacity of the covert communication channel (0.75 and 1 bit/pixel). The average values of the extraction accuracy index of additional information NCC are from 0.87 to 0.97 depending on the selected color component and the number of embedded bits. The accuracy of message extraction depends on the characteristics of the container - detection errors occur when there are a large number of blocks with brightness values close to 0 or 255. However, this problem can be eliminated by preliminary analysis of the container.The developed method showed high resistance to attack by the noise "Salt and pepper" (d=0.01) at appreciable distortions of the filled container, as well as the imposition of a Gaussian and a multiplicative noise at imperceptible distortions of stegos.