Scholarlink

<p>In recent years, technological advancements have led to improvements in wireless transmission strategies for transmitting information over the Internet. In the Medical domain, especially E-health strategies provide an effective platform for exchanging patient health information in the format of Medical images, signatures, specialized statements, and reports between other parties. In this process, medical images are sometimes vulnerable to security attacks, information leakage, alteration, and theft if they are not adequately secured. To address this limitation, this research introduces a novel zero-watermarking process for enhancing the security of medical data. Initially, the Josephus theory and dynamic cross-diffusion algorithm are used to encrypt secret images, thereby improving privacy and confidentiality. Then, absolute embedding employs a modified transformer-based encoder with a self-attention model, which extracts valuable and enriched features. These features improve performance and reduce complexity analysis. The Partial Reinforcement Optimization Algorithm is employed to tune the hyperparameter values of the modified transformer model, thereby reducing computational requirements and increasing convergence. In this proposed model, the Non-Subsampled Shearlet Transform is used to select valuable coefficient features for constructing the selected features. Finally, the XORed process combines the encrypted image with coefficient features, resulting in a zero-watermarking image that prevents unauthorized access to secret data and achieves significant performance against geometric and noise attacks, demonstrating greater robustness, imperceptibility, and versatility. In experimental analysis, the proposed model achieves peak signal-to-noise ratio (PSNR) values of 57.80, structural similarity index measurement (SSIM) of 0.9998, and bit error rate (BER) values of 0.0004, which compare favorably with the existing model, demonstrating its higher performance, invisibility, and robustness. This proposed model utilizes large-scale medical datasets and computationally intensive attention processes, necessitating high-performance computing (HPC). HPC process facilitate effective and scalable training processing and decrease computational complexity.</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

AEMT-ZM: absolute embedding enclosed modified transformers with self-attention based zero-watermarking for securing medical data

  • 期刊论文

  • 发表日期：

    2026年4月16日

查看全文
  • Dhiran Kumar Mahto,
  • Rajesh Dwivedi

摘要

In recent years, technological advancements have led to improvements in wireless transmission strategies for transmitting information over the Internet. In the Medical domain, especially E-health strategies provide an effective platform for exchanging patient health information in the format of Medical images, signatures, specialized statements, and reports between other parties. In this process, medical images are sometimes vulnerable to security attacks, information leakage, alteration, and theft if they are not adequately secured. To address this limitation, this research introduces a novel zero-watermarking process for enhancing the security of medical data. Initially, the Josephus theory and dynamic cross-diffusion algorithm are used to encrypt secret images, thereby improving privacy and confidentiality. Then, absolute embedding employs a modified transformer-based encoder with a self-attention model, which extracts valuable and enriched features. These features improve performance and reduce complexity analysis. The Partial Reinforcement Optimization Algorithm is employed to tune the hyperparameter values of the modified transformer model, thereby reducing computational requirements and increasing convergence. In this proposed model, the Non-Subsampled Shearlet Transform is used to select valuable coefficient features for constructing the selected features. Finally, the XORed process combines the encrypted image with coefficient features, resulting in a zero-watermarking image that prevents unauthorized access to secret data and achieves significant performance against geometric and noise attacks, demonstrating greater robustness, imperceptibility, and versatility. In experimental analysis, the proposed model achieves peak signal-to-noise ratio (PSNR) values of 57.80, structural similarity index measurement (SSIM) of 0.9998, and bit error rate (BER) values of 0.0004, which compare favorably with the existing model, demonstrating its higher performance, invisibility, and robustness. This proposed model utilizes large-scale medical datasets and computationally intensive attention processes, necessitating high-performance computing (HPC). HPC process facilitate effective and scalable training processing and decrease computational complexity.