开放环境下自适应聚类优化包络的相机来源取证

doi:10.12263/DZXB.20211428

电子学报 ›› 2022, Vol. 50 ›› Issue (8): 1840-1850.DOI: 10.12263/DZXB.20211428

开放环境下自适应聚类优化包络的相机来源取证

王波¹, 王悦¹, 王伟², 侯嘉尧¹

^1.大连理工大学信息与通信工程学院，辽宁大连 116024
^2.中国科学院自动化所智能感知与计算研究中心，北京 100190

收稿日期:2021-10-19 修回日期:2022-02-09 出版日期:2022-08-25
- 通讯作者:
- 王伟
- 作者简介:
- 王波男，1981年生于四川省自贡市.大连理工大学信息与通信工程学院副教授，博士生导师.主要研究方向为人工智能安全和多媒体信息安全.E-mail: bowang@dlut.edu.cn
  王悦女，1997年生于黑龙江省哈尔滨市.大连理工大学信息与通信工程学院硕士研究生.主要研究方向为数字图像来源鉴别.E-mail： yuewang9779@163.com
  王伟男,1984年生于山西省忻州市.中国科学院自动化所副研究员.研究方向多媒体内容安全、人工智能安全、多模态内容分析与理解.E-mail: wei.wong@ia.ac.cn
  侯嘉尧男,1998年生于辽宁省开原市.大连理工大学信息与通信工程学院硕士研究生.主要研究方向为小样本情况下的相机来源鉴别.E-mail: 32009157@mail.dlut.edu.cn
- 基金资助:
- 国家自然科学基金 (U1936117); 大连市科技创新基金 (2021JJ12GX018); 模式识别国家重点实验室开放课题基金 (202100032); 中央高校基本科研业务费专项资金 (No.DUT21GF303 (DUT20TD110); No.DUT20RC (3）088

Envelope Optimization Based on Adaptive Clustering for Open-Set Camera Model Identification

WANG Bo¹, WANG Yue¹, WANG Wei², HOU Jia-yao¹

^1.Faculty of Electronic Information and Electrical Engineering，Dalian University of Technology，Dalian，Liaoning 116024，China
^2.Center for Research on Intelligent Perception and Computing （CRIPAC），Institute of Automation，Chinese Academy of Sciences，Beijing 100190，China

Received:2021-10-19 Revised:2022-02-09 Online:2022-08-25 Published:2022-09-08
- Corresponding author:
- WANG Wei

摘要/Abstract

摘要：

针对相机来源取证中的开放环境问题，本文提出一种自适应聚类优化包络的相机来源取证方法，解决了现有方法在训练相机模型数量少的恶劣情况下检测精度低的问题.首先，通过手肘法得到每一类相机数据的聚类个数，并以该聚类数为参照进行 $k$ -means聚类；然后将得到的相机模型子类数据分别进行支持向量数据描述以刻画其子包络，并根据所属相机模型类别将子包络合成一个更具细节特征的特征包络；最后通过判决法则将来自未知相机模型的图像排除，并将判断为已知来源的图像分类溯源，进而实现开放环境下的相机来源鉴别.实验结果表明，在Dresden和SOCRatES两个公开数据集上，本文提出的算法具有更优的鲁棒性和扩展性，与已有方法相比，在KACC，UACC和OACC三个评估指标和时间复杂度上均表现出更优越的性能.

关键词: 开放环境, 数字图像取证, 相机来源鉴别, 手肘法, 聚类包络优化

Abstract:

In this paper, an envelope optimization based on adaptive clustering for open-set camera model identification is proposed for the open-set problem of source camera identification, which solved the problem of low detection accuracy of the existing methods in the bad situation with few known camera models. Firstly, the clustering number of each type of camera data is obtained by the elbow method, and k-means clustering is performed with this clustering number as the reference. Then the sub-class data of the camera model are described by the technique of support vector data description, respectively to describe its hypersphere sub-envelope, and the sub-envelope is synthesized into a new hypersphere envelope with more detailed features according to the class of the camera model. Finally, the images from unknown camera models are excluded by the decision rule, and the images from known sources are classified and traceable to achieve source camera identification in the open-set. Experimental results on the two public datasets Dresden and SOCRatES show that the method proposed in this paper has better robustness and scalability. Compared with the existing methods, the three evaluation indicators of KACC, UACC, and OACC and time complexity are superior.

Key words: open-set, digital image forensics, source camera identification, elbow method, envelope of clustering optimization

中图分类号:

TP391

王波, 王悦, 王伟, 侯嘉尧. 开放环境下自适应聚类优化包络的相机来源取证[J]. 电子学报, 2022, 50(8): 1840-1850.

WANG Bo, WANG Yue, WANG Wei, HOU Jia-yao. Envelope Optimization Based on Adaptive Clustering for Open-Set Camera Model Identification[J]. Acta Electronica Sinica, 2022, 50(8): 1840-1850.

图/表 18

参考文献 31

1	LI Chang-tsun. Source camera identification using enhanced sensor pattern noise[J]. IEEE Transactions on Information Forensics and Security, 2010, 5(2): 280-287.
2	BAYRAM S, SENCAR H, MEMON N, et al. Source camera identification based on CFA interpolation[C]//IEEE International Conference on Image Processing. Genoa: IEEE, 2005, 3: III-69.
3	AHMED F, KHELIFI F, LAWGALY A, et al. comparative analysis of a deep convolutional neural network for source camera identification[C]//2019 IEEE 12th International Conference on Global Security, Safety and Sustainability. London: IEEE, 2019: 1-6.
4	YANG Peng-peng, NI Rong-rong, ZHAO Yao, et al. Source camera identification based on content-adaptive fusion residual networks[J]. Pattern Recognition Letters, 2019, 119: 195-204.
5	LI R Z, LI C T, GUAN Y. Inference of a compact representation of sensor fingerprint for source camera identification[J]. Pattern Recognition, 2018, 74: 556-567.
6	KANG X G, LI Y X, QU Z H, et al. Enhancing source camera identification performance with a camera reference phase sensor pattern noise[J]. IEEE Transactions on Information Forensics and Security, 2012, 7(2): 393-402.
7	CHOI K S, LAM E Y, WONG K K Y. Automatic source camera identification using the intrinsic lens radial distortion[J]. Optics Express, 2006, 14(24): 11551-11565.
8	ZHENG Y, CAO Y, CHANG C H. A PUF-based data-device hash for tampered image detection and source camera identification[J]. IEEE Transactions on Information Forensics and Security, 2020, 15: 620-634.
9	SCHWEIGHOFER G, SEGVIC S, PINZ A. Online/realtime structure and motion for general camera models[C]//2008 IEEE Workshop on Applications of Computer Vision. Copper Mountain: IEEE, 2008: 1-6.
10	乔通, 姚宏伟, 潘彬民, 等. 基于深度学习的数字图像取证技术研究进展[J]. 网络与信息安全学报, 2021, 7(5): 13-28.
	QIAO T, YAO H W, PAN B M, et al. Research progress of digital image forensic techniques based on deep learning[J]. Chinese Journal of Network and Information Security, 2021, 7(5): 13-28. (in Chinese)
11	王波, 孔祥维, 付海燕. 联合OC-SVM和MC-SVM的图像来源取证方法[J]. 计算机研究与发展, 2009, 46(9): 1456-1461.
	WANG B, KONG X W, FU H Y. A source camera identification method based on the combination of OC-SVM and MC-SVM[J]. Journal of Computer Research and Development, 2009, 46(9): 1456-1461. (in Chinese)
12	COSTA F D O, ECKMANN M, SCHEIRER W J, et al. Open set source camera attribution[C]//2012 25th SIBGRAPI Conference on Graphics, Patterns and Images. Ouro Preto: IEEE, 2012: 71-78.
13	DE O COSTA F, SILVA E, ECKMANN M, et al. Open set source camera attribution and device linking[J]. Pattern Recognition Letters, 2014, 39: 92-101.
14	HUANG Y G, ZHANG J, HUANG H Y. Camera model identification with unknown models[J]. IEEE Transactions on Information Forensics and Security, 2015, 10(12): 2692-2704.
15	LEKSHMI K, VAITHIYANATHAN V. Source camera identification of image for forensic analysis using sensor fingerprints[C]//2018 Fourth International Conference on Computing Communication Control and Automation(ICCUBEA). Pune: IEEE, 2018: 1-5.
16	HU W P, QIN Q, WANG M Y, et al. Continual Learning by Using Information of Each Class Holistically[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35(9) 7797-7805.
17	BONDI L, GÜERA D, BAROFFIO L, et al. A preliminary study on convolutional neural networks for camera model identification[J]. Media Watermarking, Security, and Forensics, 2017, 29: 67-76.
18	BAYAR B, STAMM M C. Towards open set camera model identification using a deep learning framework[C]//2018 IEEE International Conference on Acoustics, Speech and Signal Processing. Calgary: IEEE, 2018: 2007-2011.
19	MAYER O, STAMM M C. Learned forensic source similarity for unknown camera models[C]//2018 IEEE International Conference on Acoustics, Speech and Signal Processing. Calgary: IEEE, 2018: 2012-2016.
20	MAYER O, HOSLER B, STAMM M C. Open set video camera model verification[C]//ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing. Barcelona: IEEE, 2020: 2962-2966.
21	HUMAIRA H, RASYIDAH R. Determining the appropiate cluster number using elbow method for K-means algorithm[C]//Proceedings of the 2nd Workshop on Multidisciplinary and Applications(WMA 2018). Padang: EAI, 2018: 24-25
22	HUANG Z X. Extensions to the k-means algorithm for clustering large data sets with categorical values[J]. Data Mining and Knowledge Discovery, 1998, 2: 283-304.
23	TAX D M J, DUIN R P W. Support vector data description[J]. Machine Learning, 2004, 54: 45-66.
24	LIU B, XIAO Y S, CAO L B, et al. SVDD-based outlier detection on uncertain data[J]. Knowledge and Information Systems, 2013, 34: 597-618.
25	王悦, 王波. 一种基于包络优化的图像来源鉴别方法: CN112418348A[P]. 2021-02-26.
	WANG Y, WANG B. Image source identification method based on envelope optimization: CN112418348A[P]. 2021-02-26.(in Chinese).
26	LÜCKE J, FORSTER D. k-means as a variational EM approximation of Gaussian mixture models[J]. Pattern Recognition Letters, 2019, 125: 349-356.
27	WANG B, ZHONG K, SHAN Z H, et al. A unified framework of source camera identification based on features[J]. Forensic Science International, 2020, 307: 110109.
28	ZHANG G W, WANG B, WEI F, et al. Source camera identification for re-compressed images: A model perspective based on tri-transfer learning[J]. Computers & Security, 2021, 100: 102076.
29	WANG B, WANG Y, HOU J Y, et al. Discriminative feature projection for camera model identification of recompressed images[J]. Multimedia Tools and Applications, 2021, 80: 29719-29743.
30	GLOE T, BÖHME R. The Dresden image database for benchmarking digital image forensics[J]. Journal of Digital Forensic Practice, 2010, 3(2/3/4): 150-159.
31	GALDI C, HARTUNG F, DUGELAY J L. SOCRatES: A database of realistic data for SOurce camera REcognition on smartphones[C]//Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods. Prague: SciTePress, 2019: 648-655.

相机模型	缩写	数量	尺寸
Kodak_M1063	K1	2 391	3 664×2 748
Olympus_mju_1050SW	O1	1 037	3 648×2 736
Praktica_DCZ5_9	P1	1 019	2 560×1 920
Panasonic_DMC-FZ50	Pa1	931	3 648×2 736
Casio_EX-Z150	C1	925	3 264×2 448
Nikon_CoolPixS710	N1	925	4 352×3 264
Ricoh_GX100	R1	854	3 648×2 736
Nikon_D200	N2	752	3 872×2 592
Sony_DSC-T77	S1	725	3 648×2 736
Samsung_L74wide	Sa1	686	3 072×2 304

相机模型	缩写	数量	尺寸
Kodak_M1063	K1	2 391	3 664×2 748
Olympus_mju_1050SW	O1	1 037	3 648×2 736
Praktica_DCZ5_9	P1	1 019	2 560×1 920
Panasonic_DMC-FZ50	Pa1	931	3 648×2 736
Casio_EX-Z150	C1	925	3 264×2 448
Nikon_CoolPixS710	N1	925	4 352×3 264
Ricoh_GX100	R1	854	3 648×2 736
Nikon_D200	N2	752	3 872×2 592
Sony_DSC-T77	S1	725	3 648×2 736
Samsung_L74wide	Sa1	686	3 072×2 304

相机模型	缩写	数量	尺寸
Apple iPhone 5s	A1	180	3 264×2 488
Apple iPhone 6	A2	180	3 264×2 488
Apple iPhone 6s	A3	180	4 032×3 024
Apple iPhone 7	A4	180	4 032×3 024
Asus Zenfone 2	AZ1	180	4 096×3 072
LG G3	L1	180	4 160×2 340
Motoroal Moto G	M1	180	3 264×1 836
Samsung Galaxy A3	S1	180	4 128×2 322
Samsung Galaxy S5	S2	180	5 312×2 988
Samsung Galaxy S7 Edge	S3	180	4 032×3 024

相机模型	缩写	数量	尺寸
Apple iPhone 5s	A1	180	3 264×2 488
Apple iPhone 6	A2	180	3 264×2 488
Apple iPhone 6s	A3	180	4 032×3 024
Apple iPhone 7	A4	180	4 032×3 024
Asus Zenfone 2	AZ1	180	4 096×3 072
LG G3	L1	180	4 160×2 340
Motoroal Moto G	M1	180	3 264×1 836
Samsung Galaxy A3	S1	180	4 128×2 322
Samsung Galaxy S5	S2	180	5 312×2 988
Samsung Galaxy S7 Edge	S3	180	4 032×3 024

Dresden		SOCRatES
相机模型	个数	相机模型	个数
K1	2	A1	1
O1	2	A2	1
P1	3	A3	1
Pa1	2	A4	2
C1	2	AZ1	1
N1	2	L1	1
R1	1	M1	2
N2	2	S1	1
S1	2	S2	1
Sa1	2	S3	2

开放环境下自适应聚类优化包络的相机来源取证

Envelope Optimization Based on Adaptive Clustering for Open-Set Camera Model Identification

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 18

参考文献 31

相关文章 15

编辑推荐

Metrics

相机模型	算法
相机模型	SCIU	CCF	CCF_AC	SC	本文算法
K1	52.0	51.1	60.2	93.2	94.4
O1	48.3	42.8	47.3	87.9	95.1
P1	44.0	45.1	50.6	86.0	91.2
Pa1	36.3	44.8	52.5	93.0	98.2
C1	67.4	45.8	51.7	93.1	94.1
N1	34.1	43.1	53.5	86.6	89.9
R1	76.6	50.8	50.8	100.0	100.0
N2	23.3	51.9	53.5	90.1	96.3
S1	34.1	42.1	56.7	97.1	98.4
Sa1	12.8	47.2	56.1	88.2	94.1
AVG	42.9	46.5	53.3	91.5	95.2

相机模型	SCIU	CCF	CCF_AC	SC	本文算法
K1	18 703.8	948.2	375.2	11.2	8.1
O1	14 540.3	111.8	48.3	6.9	5.2
P1	13 957.0	103.1	57.6	9.0	5.3
Pa1	14 397.3	36.8	78.5	6.3	9.9
C1	16 347.3	95.8	34.7	6.2	9.8
N1	13 573.1	77.1	34.5	6.4	10.0
R1	16 283.2	72.8	72.8	6.3	6.3
N2	16 367.3	47.9	22.5	6.2	9.1
S1	12 847.1	45.1	24.7	6.0	9.4
Sa1	10 237.8	38.2	19.1	6.2	9.6
AVG	14 725.4	157.7	76.8	7.1	8.3

相机模型	SCIU	CCF	CCF_AC	SC	本文算法
A1	8 162.7	397.0	397.0	9.2	9.2
A2	8 825.3	408.1	408.1	8.7	8.7
A3	8 901.7	414.8	414.8	9.6	9.6
A4	9 174.7	412.3	9.1	9.7	8.9
AZ1	9 145.6	441.3	441.3	8.7	8.7
L1	9 318.7	391.5	391.5	9.0	9.0
M1	9 134.9	413.5	11.4	9.8	8.5
S1	8 932.7	451.5	451.5	8.9	8.9
S2	8 143.7	419.5	419.5	8.8	8.8
S3	8 923.9	493.4	12.5	8.7	7. 9
AVG	8 866.4	424.3	295.7	9.1	8.8

[1]	王子为, 鲁继文, 周杰. 基于自适应梯度优化的二值神经网络[J]. 电子学报, 2023, (): 1-10.
[2]	刘金平, 吴娟娟, 张荣, 徐鹏飞. 基于结构重参数化与多尺度深度监督的COVID-19胸部CT图像自动分割[J]. 电子学报, 2023, (): 1-9.
[3]	张笑宇, 沈超, 蔺琛皓, 李前, 王骞, 李琦, 管晓宏. 面向机器学习模型安全的测试与修复[J]. 电子学报, 2023, (): 1-35.
[4]	王炼红, 罗志辉, 林飞鹏, 李潇瑶. 采用多头注意力机制的C&RM-MAKT预测算法[J]. 电子学报, 2022, (): 1-9.
[5]	苏田田, 王慧敏, 张小凤. 基于多分支瓶颈结构的轻量型图像分类算法研究[J]. 电子学报, 2022, (): 1-9.
[6]	刘芳, 朱天贺, 苏卫星, 刘阳. 基于高斯隐马尔可夫模型的人机共享控制区域化决策算法[J]. 电子学报, 2022, 50(11): 2659-2667.
[7]	桑海峰, 陈旺兴, 王海峰, 王金玉. 基于多模式时空交互的行人轨迹预测模型[J]. 电子学报, 2022, 50(11): 2806-2812.
[8]	刘耿耿, 李泽鹏, 郭文忠, 陈国龙, 徐宁. 面向超大规模集成电路物理设计的通孔感知的并行层分配算法[J]. 电子学报, 2022, 50(11): 2575-2583.
[9]	魏博文, 全红艳. 基于语义与形态特征融合的语义分割网络[J]. 电子学报, 2022, 50(11): 2688-2697.
[10]	姚睿, 朱享彬, 周勇, 王鹏, 张艳宁, 赵佳琦. 基于重要特征的视觉目标跟踪可迁移黑盒攻击方法[J]. 电子学报, 2022, (): 1-10.
[11]	金紫凤, 潘思聪, 危辉. 可变环境下基于位姿变换矩阵的机器人无标定手眼协调方法[J]. 电子学报, 2022, 50(10): 2318-2328.
[12]	魏钰轩, 陈莹. 基于自适应层信息熵的卷积神经网络压缩[J]. 电子学报, 2022, 50(10): 2398-2408.
[13]	马百腾, 张士伟, 高常鑫, 桑农. 面向行为边界框生成的端到端时间全局相关网络[J]. 电子学报, 2022, 50(10): 2452-2461.
[14]	肖斌, 陈嘉博, 毕秀丽, 张俊辉, 李伟生, 王国胤, 马旭. 基于一维卷积神经网络与循环神经网络串联的心音分析方法[J]. 电子学报, 2022, 50(10): 2425-2432.
[15]	周登文, 李文斌, 李金新, 黄志勇. 一种轻量级的多尺度通道注意图像超分辨率重建网络[J]. 电子学报, 2022, 50(10): 2336-2346.

相机模型	算法
相机模型	SCIU	CCF	CCF_AC	SC	本文算法
K1	78.3	85.6	92.7	84.1	87.2
O1	80.0	95.0	100.0	97.4	97.6
P1	75.6	97.1	96.2	94.9	95.8
Pa1	84.2	91.4	93.9	97.1	96.9
C1	87.5	99.6	99.6	94.4	96.4
N1	93.5	98.1	97.8	96.9	97.0
R1	81.8	99.0	99.0	98.8	98.8
N2	85.3	100.0	100.0	99.8	99.8
S1	92.0	93.9	94.0	95.4	96.9
Sa1	89.5	98.0	98.0	94.6	95.1
AVG	84.8	95.8	97.1	95.3	96.1

相机模型	算法
相机模型	SCIU	CCF	CCF_AC	SC	本文算法
K1	78.3	85.6	92.7	84.1	87.2
O1	80.0	95.0	100.0	97.4	97.6
P1	75.6	97.1	96.2	94.9	95.8
Pa1	84.2	91.4	93.9	97.1	96.9
C1	87.5	99.6	99.6	94.4	96.4
N1	93.5	98.1	97.8	96.9	97.0
R1	81.8	99.0	99.0	98.8	98.8
N2	85.3	100.0	100.0	99.8	99.8
S1	92.0	93.9	94.0	95.4	96.9
Sa1	89.5	98.0	98.0	94.6	95.1
AVG	84.8	95.8	97.1	95.3	96.1

相机模型	算法
相机模型	SCIU	CCF	CCF_AC	SC	本文算法
K1	72.2	77.5	85.0	86.2	88.8
O1	76.8	89.6	94.6	96.4	97.3
P1	72.4	91.9	91.6	94.0	95.3
Pa1	79.8	87.1	90.1	96.7	97.0
C1	85.6	94.6	95.2	94.3	96.2
N1	88.1	93.1	93.7	95.9	96.3
R1	81.3	94.9	94.9	98.9	98.9
N2	80.7	96.4	96.6	99.1	99.5
S1	87.9	90.2	91.3	95.6	97.0
Sa1	84.3	94.6	95.2	94.1	95.0
AVG	80.9	91.0	92.8	95.1	96.1

相机模型	算法
相机模型	SCIU	CCF	CCF_AC	SC	本文算法
K1	72.2	77.5	85.0	86.2	88.8
O1	76.8	89.6	94.6	96.4	97.3
P1	72.4	91.9	91.6	94.0	95.3
Pa1	79.8	87.1	90.1	96.7	97.0
C1	85.6	94.6	95.2	94.3	96.2
N1	88.1	93.1	93.7	95.9	96.3
R1	81.3	94.9	94.9	98.9	98.9
N2	80.7	96.4	96.6	99.1	99.5
S1	87.9	90.2	91.3	95.6	97.0
Sa1	84.3	94.6	95.2	94.1	95.0
AVG	80.9	91.0	92.8	95.1	96.1