Image Sterilization through Adaptive Noise Blending in Integer Wavelet Transformation
Sreeparna Ganguly, Imon Mukherjee · 2022 IEEE 19th India Council International Conference (INDICON) · 2022
Detecting steganography can be more resource-consuming than corrupting the stego-data concealed inside a cover media. Based on this fact, the current work proposes an Integer Wavelet Transformation (IWT) based image sterilization technique that corrupts text based secret data hidden inside image cover. The technique infuses random noise into the IWT sub-bands of a given cover image in such way that the image quality and visual perceptibility is maintained beyond the thresholds of Human Visual System (HVS). According to the principle of wavelet decomposition, the noise susceptibility of the IWT sub-bands differ to an wide extent. Inducing same amount of noise in all sub-bands can lead to visible degradation of the sterilization output. To mitigate this issue, the proposed work employs adaptive noise injection depending on the sub-bands being attacked. Partial bit-flipping noise is used for the most noise sensitive approximation sub-band. Moderate impact noise is infused into vertical and horizontal detail bands. The most noise susceptible diagonal detail band is corrupted with high impact noise blending. The proposed method, when applied on stegograms generated by seven popular steganography methods, is able to achieve up to 88.61% stego-clean images with average PSNR of 46.08 dB and average SSIM of 0.99.