Scholarlink

<p>Semantic communication enriches transmission efficiency, by sending only data meaning. However, unlike modern communications that provide bit-level security, the eavesdropping of semantic communications poses a significant risk of potentially exposing the intention of legitimate user during semantic-rich data transmission over insecure channels that introduces risks in privacy and security. This paper addresses the privacy issue when transmitting images over wireless channels in the presence of illegitimate receivers and proposes a novel semantic communication approach ensuring privacy through a dual subsequence of obfuscation and security layers, and feature transformation for image protection and deprotection mechanisms. The proposed system exploits the potesntial of Channel and Spatial Attention mechanisms for nodes optimization and utilizes the joint-Gaussian Differential Privacy and Feature Transformation technique, referred to as CSA–GDPFT, to secure semantic information routed from the high-altitude platform station (HAPS) node to the ground (terrestrial) legitimate receiver. Simulation results confirm the effectiveness of the proposed system in protecting ground illegitimate receivers from obtaining semantic information while maintaining high–fidelity image reconstruction at the legitimate receiver. By using the proposed scheme, the legitimate receiver demonstrates an ability to achieve a peak signal-to-noise ratio (PSNR) of more than 24&#xa0;dB and a learned perceptual image patch similarity (LPIPS) of less than 0.075 when evaluated using CelebAMask–HQ dataset over 200 epochs and various privacy budget values. The PSNR of nearly 30&#xa0;dB and an LPIPS less than 0.025 can be achieved when the proposed scheme is evaluated using both of CIFAR-10 and CIFAR-100 datasets. On the other hand, the illegitimate receiver by using its architecture can achieve an almost constant PSNR of 11&#xa0;dB and an LPIPS of around 0.78 which demonstrates its inability to successfully reconstruct the received message. Furthermore, the proposed system reveals to maintain reliable performance in various privacy settings.</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

Attention–based HAPS–to–ground nodes optimization for differential privacy towards secure semantic communications

  • 期刊论文

  • 发表日期：

    2026年3月4日

查看全文
  • Pascal Nkurunziza,
  • Daisuke Umehara

摘要

Semantic communication enriches transmission efficiency, by sending only data meaning. However, unlike modern communications that provide bit-level security, the eavesdropping of semantic communications poses a significant risk of potentially exposing the intention of legitimate user during semantic-rich data transmission over insecure channels that introduces risks in privacy and security. This paper addresses the privacy issue when transmitting images over wireless channels in the presence of illegitimate receivers and proposes a novel semantic communication approach ensuring privacy through a dual subsequence of obfuscation and security layers, and feature transformation for image protection and deprotection mechanisms. The proposed system exploits the potesntial of Channel and Spatial Attention mechanisms for nodes optimization and utilizes the joint-Gaussian Differential Privacy and Feature Transformation technique, referred to as CSA–GDPFT, to secure semantic information routed from the high-altitude platform station (HAPS) node to the ground (terrestrial) legitimate receiver. Simulation results confirm the effectiveness of the proposed system in protecting ground illegitimate receivers from obtaining semantic information while maintaining high–fidelity image reconstruction at the legitimate receiver. By using the proposed scheme, the legitimate receiver demonstrates an ability to achieve a peak signal-to-noise ratio (PSNR) of more than 24 dB and a learned perceptual image patch similarity (LPIPS) of less than 0.075 when evaluated using CelebAMask–HQ dataset over 200 epochs and various privacy budget values. The PSNR of nearly 30 dB and an LPIPS less than 0.025 can be achieved when the proposed scheme is evaluated using both of CIFAR-10 and CIFAR-100 datasets. On the other hand, the illegitimate receiver by using its architecture can achieve an almost constant PSNR of 11 dB and an LPIPS of around 0.78 which demonstrates its inability to successfully reconstruct the received message. Furthermore, the proposed system reveals to maintain reliable performance in various privacy settings.