Multilayer image authentication using deep search convolution for image integrity protection

Abstract

A fragile image watermarking framework can be employed to ascertain any modifications that may have occurred within the image content. This manuscript introduces a deep search convolution for the localization of tampering, utilizing multiple layers of authentication alongside a chaotic map. The proposed approach generates three distinct metrics of scrambled bits through the application of a chaotic map. Each matrix corresponds in size to the cover color image. Subsequently, the watermark data is created using the parity of the seven most significant bits of the image, along with a scrambled bit, and then embedded into the least significant bit of each pixel. The findings illustrate that our scheme is capable of accurately identifying tampered regions in scenarios involving copy-move forgery, removal, additional text, noise, and collage attacks. The proposed scheme attained a remarkable tamper localization accuracy of approximately 0.9995, alongside an average computational time of about 6.5964 s, which is superior to or comparable with existing tamper detection algorithms. Furthermore, the proposed method yielded a PSNR of 51 dB and an SSIM of 0.9992.