一种加密遥感图像的安全外包搜索方案 下载: 902次
A Secure Outsourcing Search Scheme for Encrypted Remote Sensing Images
1 上海海洋大学信息学院, 上海 201306
2 国家海洋局东海分局, 上海 200136
图 & 表
图 1. 加密遥感图像的安全外包搜索方案系统模型
Fig. 1. System model for secure outsourcing search for encrypted remote sensing images
下载图片 查看原文
图 2. 图像持有者上传图像流程图
Fig. 2. Flow chart of uploading image by image owner
下载图片 查看原文
图 3. 加密遥感图像的搜索方案基本流程图
Fig. 3. Basic flow chart for search of encrypted remote sensing images
下载图片 查看原文
图 4. 图像XOR加密流程图
Fig. 4. Flow chart of XOR encryption of images
下载图片 查看原文
图 5. 置换表
Fig. 5. Permutation table
下载图片 查看原文
图 6. JL转换流程图
Fig. 6. Flow chart of JL transformation
下载图片 查看原文
图 7. 二值化攻击效果图。(a)原图;(b)密文图像;(c)二值化攻击结果图像
Fig. 7. Binary attack effect. (a) Original image; (b) encrypted image; (c) binary attack result
下载图片 查看原文
图 8. 基本扫描模式
Fig. 8. Basic scan modes
下载图片 查看原文
图 9. 分划模式和变换
Fig. 9. Mode division and transformation
下载图片 查看原文
图 10. 扫描密钥B5(s2 Z0(c5 b0 o0 s5) c4 d1)路径示意图
Fig. 10. Path schematic for scanning of key B5(s2 Z0(c5 b0 o0 s5) c4 d1)
下载图片 查看原文
图 11. 遥感图像搜索结果。(a)图像O;(b)不同时相的图像;(c)含高斯分布噪声的遥感图像;(d)光线暗沉的遥感图像;(e)发生形变的遥感图像;(f)岛屿的遥感图像;(g)陆地的遥感图像
Fig. 11. Search results of remote sensing images. (a) Image O; (b) image with different phases; (c) remote sensing image with Gaussian noise; (d) remote sensing image under dimmed light; (e) remote sensing image with deformation; (f) remote sensing image of island; (g) remote sensing image of land
下载图片 查看原文
图 12. 查准率对比图
Fig. 12. Comparison of precision rates
下载图片 查看原文
图 13. 查全率对比图
Fig. 13. Comparison of recall rates
下载图片 查看原文
图 14. 搜索时间对比
Fig. 14. Comparison of search time
下载图片 查看原文
图 15. 二值化攻击效果图。(a)基本方案;(b)改进方案
Fig. 15. Binary attack effect. (a) Basic scheme; (b) improved scheme
下载图片 查看原文
图 16. 原始图像中相邻像元间的相关关系
Fig. 16. Correlation among adjacent pixels in original image
下载图片 查看原文
图 17. 加密图像中相邻像元间的相关关系
Fig. 17. Correlation among adjacent pixels in encrypted image
下载图片 查看原文
表 1算法1遥感图像加密算法
Table1. Algorithm 1 of encryption algorithm for remote sensing images
Input | Split an index remote sensing image band into a single-band image Y(i,j);Height and width of remote sensing image are H and W;Randomly generate a matrix Q of Gaussian distribution h2×k obeying an average of 0 and a variance of 1/k;Randomly generate a matrix R of H×W;Randomly generate a matrix Δ of h2×k; |
---|
Ouput | XOR encrypted remote sensing image EA;JL transformation encrypted remote sensing image EB;Double ciphertext EB={EA,EB} | Algorithmicprocess | 1. for image Y i=1 to H {2. for image Y j=1 to W {3. EA(i,j)=Y(i,j)⊕R(i,j)4. }5. }6. for image Y i=1 to h2{7. for image Y j=1 to h2{8. for matrix Q m=1 to h2{9. for matrix Q n=1 to k {10. EB(m,n)=Y(i,j)×Q(m,n)+Δ(m,n)11. }12. }13. }14. }15. return upload index EI={EA,EB} |
|
查看原文
表 2算法2遥感图像间欧氏距离计算算法
Table2. Algorithm 2 of algorithm for calculation of Euclidean distance between remote sensing images
Input | Height and width of remote sensing image EB by JL transformation are H and W; Remote sensing image S stored in cloud platform; |
---|
Ouput | Distance of two remote sensing images is d; | Algorithmicprocess | 1. for image EB i=1 to H {2. for image EB j=1 to W {3. sum+=(EB(i,j)-S(i,j))24. }5. }6. d=7. return distance d |
|
查看原文
表 3最优阈值分析表
Table3. Optimal threshold analysis
(T1,T2) | Actual number ofsimilar images | Number ofimages searched | Number ofcorrect images | Number oferror images | Number of imagesnot searched |
---|
(0.7,0.8) | 56 | 0 | 0 | 0 | 56 | (0.8,0.9) | 56 | 0 | 0 | 0 | 56 | (0.9,1.0) | 56 | 4 | 4 | 0 | 52 | (1.0,1.1) | 56 | 12 | 12 | 0 | 44 | (1.1,1.2) | 56 | 52 | 32 | 20 | 24 | (1.2,1.3) | 56 | 56 | 0 | 56 | 56 | (0.8,1.0) | 56 | 4 | 4 | 0 | 52 | (0.9,1.1) | 56 | 20 | 20 | 0 | 36 | (1.0,1.2) | 56 | 64 | 44 | 20 | 12 | (0.8,1.1) | 56 | 20 | 20 | 0 | 36 | (0.9,1.2) | 56 | 72 | 52 | 20 | 4 |
|
查看原文
表 4参数取值表
Table4. Parameter value
Parameter | Instructions | Value |
---|
H | Height of remote sensing image | 512 | W | Width of remote sensing image | 512 | N | Band number of remote sensing image | 200 | h | Height of JL transformation | 5 | k | Variance parameter ofGaussian distribution Q | 18 | s | Variance parameter ofGaussian distribution Δ | 0.5 | T | Threshold | 550 |
|
查看原文
表 5查准率
Table5. Precision rate
Scheme category | Number of images |
---|
50 | 100 | 150 | 200 | 250 | 300 | 350 | 400 |
---|
Lowe's scheme [25] | 0.135 | 0.229 | 0.378 | 0.446 | 0.568 | 0.608 | 0.676 | 0.729 | Basic scheme | 0.059 | 0.101 | 0.166 | 0.196 | 0.25 | 0.268 | 0.304 | 0.333 | Improved scheme | 0.139 | 0.222 | 0.375 | 0.431 | 0.556 | 0.597 | 0.667 | 0.722 |
|
查看原文
表 6查全率
Table6. Recall rate
Scheme category | Number of images |
---|
50 | 100 | 150 | 200 | 250 | 300 | 350 | 400 |
---|
Lowe's scheme [25] | 0.179 | 0.304 | 0.500 | 0.589 | 0.750 | 0.804 | 0.893 | 0.964 | Basic scheme | 0.179 | 0.304 | 0.500 | 0.589 | 0.750 | 0.804 | 0.911 | 1 | Improved scheme | 0.179 | 0.286 | 0.482 | 0.554 | 0.714 | 0.768 | 0.857 | 0.929 |
|
查看原文
表 7原始图像与加密图像的相邻像素点数据表
Table7. Adjacent pixel point data in original and encrypted images
Original image | Pixel value of point(x, y) | 156 | 139 | 143 | 138 | 139 | 147 | 153 | 156 | 157 | 146 |
---|
Pixel value of point (x+1, y) | 77 | 57 | 71 | 56 | 48 | 49 | 61 | 78 | 81 | 91 |
---|
Encrypted image | Pixel value of point(x,y) | 156 | 139 | 143 | 138 | 139 | 147 | 153 | 156 | 151 | 156 | Pixel value of point (x+1, y) | 33 | 111 | 19 | 63 | 47 | 114 | 24 | 88 | 37 | 32 |
|
查看原文
表 8密文搜索方案分析表
Table8. Analysis of related schemes for ciphertext search
Ciphertextsearch scheme | Suitable forcloud computing | Support imagesearch | Security | Accuracy | Searchefficiency |
---|
Ref. [26] | Yes | False | Low | Low | Low | Ref. [27] | False | False | High | High | Low | Ref. [28] | False | Yes | Low | High | Low | Ref. [13] | Yes | Yes | Higher | High | Higher | Proposed scheme | Yes | Yes | High | High | High |
|
查看原文
黄冬梅, 吴国健, 魏立斐, 魏泉苗, 戴亮. 一种加密遥感图像的安全外包搜索方案[J]. 激光与光电子学进展, 2019, 56(3): 031001. Dongmei Huang, Guojian Wu, Lifei Wei, Quanmiao Wei, Liang Dai. A Secure Outsourcing Search Scheme for Encrypted Remote Sensing Images[J]. Laser & Optoelectronics Progress, 2019, 56(3): 031001.