Scholarlink

<p>To provide a safe and decentralized means for multimedia data verification, a hybrid deep learning approach for image watermarking coupled with Blockchain technology is presented. A Wasserstein Generative Adversarial Network (WGAN) with Gradient Penalty (GP) and a Conditional GAN (cGAN) are used to implement an evolving watermarking method. A method of creating a mark that is aware of classes without being visually detectable is provided by utilising both the WGAN-GP and cGAN. The next step is to insert the embossed marks into pictorial images by creating the embossed marks using the DWT (Discrete Wavelet Transform), thereby providing a watermarking technique resistant to any type of geometric attack and signal processing. Moreover, to allow for a non-tampering means for claimant verification, a Blockchain concept is introduced to ensure that the cryptographic representation of the marked pictorial image and embossed markers is recorded on a Blockchain network through smart contracts, with pictorial image metadata preserved via an InterPlanetary File System (IPFS). An NFT is created to link ownership with a cryptocurrency wallet address for decentralized copying right enforcement and transferability. Based on privacy-preserving verification, the framework follows optional zero-knowledge proofs (ZKP) for authenticating the authenticity of the watermarks without disclosing information about the watermarks and secret keys. The experimental results show strong imperceptibility and robustness, ensuring the efficacy of the proposed framework for secure digital content protection.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

DeepBlockMark: a decentralized GAN-based image watermarking system

  • 期刊论文

  • 发表日期：

    2026年3月13日

查看全文
  • M. Subashini,
  • P. V. Ravindranath

摘要

To provide a safe and decentralized means for multimedia data verification, a hybrid deep learning approach for image watermarking coupled with Blockchain technology is presented. A Wasserstein Generative Adversarial Network (WGAN) with Gradient Penalty (GP) and a Conditional GAN (cGAN) are used to implement an evolving watermarking method. A method of creating a mark that is aware of classes without being visually detectable is provided by utilising both the WGAN-GP and cGAN. The next step is to insert the embossed marks into pictorial images by creating the embossed marks using the DWT (Discrete Wavelet Transform), thereby providing a watermarking technique resistant to any type of geometric attack and signal processing. Moreover, to allow for a non-tampering means for claimant verification, a Blockchain concept is introduced to ensure that the cryptographic representation of the marked pictorial image and embossed markers is recorded on a Blockchain network through smart contracts, with pictorial image metadata preserved via an InterPlanetary File System (IPFS). An NFT is created to link ownership with a cryptocurrency wallet address for decentralized copying right enforcement and transferability. Based on privacy-preserving verification, the framework follows optional zero-knowledge proofs (ZKP) for authenticating the authenticity of the watermarks without disclosing information about the watermarks and secret keys. The experimental results show strong imperceptibility and robustness, ensuring the efficacy of the proposed framework for secure digital content protection.