Audio-in-Image Steganography Using Analysis and Resynthesis Sound Spectrograph

Steganography is the practice of concealing secret information within a non-secret medium, such as an image or audio file, to prevent detection by unauthorised observers. Currently, there is a shortage of solutions for audio-in-image storage, with a complete absence of methods that do not rely on deep learning approaches. This is problematic because learning models often plateau at a certain level of accuracy, making further improvements difficult. Audio steganography, like other forms of steganography, is vulnerable to data malleability and requires solutions that offer customisable security features while maintaining robustness. This paper presents a novel approach for audio-in-image steganography involving the analysis and resynthesis of sound spectrographs (ARSS). This framework ensures efficient and secure data storage using a simplistic, multi-channel, and non-lossy approach to embed an audio spectrograph in a high-resolution image file. It allows for minimal pre-processing and post-processing to conceal one or more sound spectrographs within a single cover image while focusing on maintaining high fidelity in sound resynthesis during retrieval. Referred to as ASA, this proposed approach achieves a mean PSNR value of 42.54 dB for the stego-image and 45.02 dB for the reconstructed image, outperforming most existing methods while also offering high storage efficiency and infrastructure for additional security measures.