Maximizing complex features to minimize the detectability of content-adaptive steganography
De Rosal Ignatius Moses Setiadi,Supriadi Rustad,Pulung Nurtantio Andono,Guruh Fajar Shidik
DOI: https://doi.org/10.1007/s11042-024-20056-7
IF: 2.577
2024-08-18
Multimedia Tools and Applications
Abstract:Content-adaptive steganography has three rules that need to be fulfilled to maximize the performance of the cost function, thereby increasing resistance to steganalysis attacks. This study proposes a combination method of image sharpening, Fuzzy edge detection, and residual weighting to maximize the performance of the three rules. Image sharpening is found to be useful for detailing complex features as required by the complexity first rule, while fuzzy edge detection is found to be effective in maximizing the spreading second rule. The clustering third rule is also implemented by customizing the cost of the pixels outside the embedding areas. Embedding simulation is applied to a modified MiPOD method, which we named MaxMiPOD, where SRM and maxSRMd2 steganalysis trials are carried out at payloads of 0.05, 0.1, 0.2, 0.3, 0.4, and 0.5 bpp. The proposed method produces a very significant performance increase of undetectability compared to the MiPOD standard and state-of-the-art methods, especially at a payload of 0.05 bpp.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering