Scholarlink

<p>The widespread use of high-resolution remote sensing images in meteorology, geology, and national security calls for highly efficient and secure protection mechanisms. However, conventional image encryption methods often face substantial limitations when handling large-scale geospatial data, due to memory-bandwidth constraints and insufficient computational throughput. To address these bottlenecks, this paper proposes a novel Heterogeneous CPU–GPU Parallel Image Encryption Scheme (HC-PIES), which structurally integrates chaotic permutations, DNA-level operations, and cellular-automaton (CA)-based diffusion. Within the HC-PIES architecture, the CPU initially performs a global spatial permutation stage driven by a 2D-SLMM chaotic map, utilizing session-dependent parameters dynamically derived from the SHA-512 hash of the plaintext image. Subsequently, to mitigate global memory access latency and accelerate the diffusion phase, a block-based GPU parallelization strategy is introduced. Specifically, a fused GPU kernel architecture is designed to execute chaotic sequence generation and hybrid DNA-CA diffusion within the on-chip shared memory, thereby reducing memory overhead and improving parallel execution efficiency. Extensive experiments conducted on standard test datasets and remote sensing images demonstrate that the proposed HC-PIES achieves both strong security and practical computational efficiency. For a <InlineEquation ID="IEq1"><EquationSource Format="TEX">\(2048\times 2048\times 3\)</EquationSource></InlineEquation> remote sensing image, the ciphertext information entropy reaches 7.9999 bits, closely approaching the theoretical ideal. Furthermore, the measured NPCR and UACI values satisfy the mathematical expectations for 8-bit image encryption. Performance evaluation shows that the proposed implementation encrypts a&#xa0;<InlineEquation ID="IEq2"><EquationSource Format="TEX">\(256\times 256\times 3\)</EquationSource></InlineEquation>&#xa0;image in 0.0853 s on an entry-level GPU, achieving a significant acceleration ratio over a sequential CPU implementation. These results indicate that HC-PIES is highly promising for secure and real-time processing of massive remote sensing data.</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

A CPU-GPU heterogeneous parallel encryption scheme for raster remote sensing images using hybrid DNA operations and cellular automaton diffusion

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月2日

查看全文
  • Yi Huang,
  • Jianguo Dai,
  • Guoshun Zhang,
  • Xin Zhan

摘要

The widespread use of high-resolution remote sensing images in meteorology, geology, and national security calls for highly efficient and secure protection mechanisms. However, conventional image encryption methods often face substantial limitations when handling large-scale geospatial data, due to memory-bandwidth constraints and insufficient computational throughput. To address these bottlenecks, this paper proposes a novel Heterogeneous CPU–GPU Parallel Image Encryption Scheme (HC-PIES), which structurally integrates chaotic permutations, DNA-level operations, and cellular-automaton (CA)-based diffusion. Within the HC-PIES architecture, the CPU initially performs a global spatial permutation stage driven by a 2D-SLMM chaotic map, utilizing session-dependent parameters dynamically derived from the SHA-512 hash of the plaintext image. Subsequently, to mitigate global memory access latency and accelerate the diffusion phase, a block-based GPU parallelization strategy is introduced. Specifically, a fused GPU kernel architecture is designed to execute chaotic sequence generation and hybrid DNA-CA diffusion within the on-chip shared memory, thereby reducing memory overhead and improving parallel execution efficiency. Extensive experiments conducted on standard test datasets and remote sensing images demonstrate that the proposed HC-PIES achieves both strong security and practical computational efficiency. For a \(2048\times 2048\times 3\) remote sensing image, the ciphertext information entropy reaches 7.9999 bits, closely approaching the theoretical ideal. Furthermore, the measured NPCR and UACI values satisfy the mathematical expectations for 8-bit image encryption. Performance evaluation shows that the proposed implementation encrypts a \(256\times 256\times 3\) image in 0.0853 s on an entry-level GPU, achieving a significant acceleration ratio over a sequential CPU implementation. These results indicate that HC-PIES is highly promising for secure and real-time processing of massive remote sensing data.