School of Computer and Software, Nanjing University of Information Science and Technology, Ning Liu Road, No. 219, Nanjing, 210044, China
2.
Jiangsu Engineering Centre of Network Monitoring, Ning Liu Road, No. 219, Nanjing, 210044, China
3.
School of Engineering and Applied Science, The George Washington University, The Cloyd Heck Marvin Center, 800 21st Street, NW-Suite 505, Washington, DC 20052, US
Received:
19 January 2019
Accepted:
15 May 2019
Published:
31 May 2019
Computer graphic images (CGI) can be manufactured very similar to natural images (NI) by state-of-the-art algorithms in computer graphic filed. Thus, there are various identification algorithms proposed to detect CGI. However, the manipulation is complicated and difficult for an ultimate CGI against the forensic algorithms. Further, the forensics on CGI and NI made achievements in the different aspects with the encouragement of deep learning. Though the generated CGI can achieve high quality automatically by generative adversarial networks (GAN), CGI generation based on GAN is difficult to ensure that it cannot be detected by forensics. In this paper, we propose a brief and effective architecture based on GAN for preventing the generated images being detected under the forensics on CGI and NI. The adapted characteristics will make the CGI generated by GAN fools the detector and keep the end-to-end generation mode of GAN.
Citation: Qi Cui, Ruohan Meng, Zhili Zhou, Xingming Sun, Kaiwen Zhu. An anti-forensic scheme on computer graphic images and natural images using generative adversarial networks[J]. Mathematical Biosciences and Engineering, 2019, 16(5): 4923-4935. doi: 10.3934/mbe.2019248
Related Papers:
Abstract
Computer graphic images (CGI) can be manufactured very similar to natural images (NI) by state-of-the-art algorithms in computer graphic filed. Thus, there are various identification algorithms proposed to detect CGI. However, the manipulation is complicated and difficult for an ultimate CGI against the forensic algorithms. Further, the forensics on CGI and NI made achievements in the different aspects with the encouragement of deep learning. Though the generated CGI can achieve high quality automatically by generative adversarial networks (GAN), CGI generation based on GAN is difficult to ensure that it cannot be detected by forensics. In this paper, we propose a brief and effective architecture based on GAN for preventing the generated images being detected under the forensics on CGI and NI. The adapted characteristics will make the CGI generated by GAN fools the detector and keep the end-to-end generation mode of GAN.
References
[1]
A. Alabdulkarim, M. Al-Rodhaan, Y. Tian, et al., A privacy-preserving algorithm for clinical decision-support systems using random forest, bCMC-Comput. Mater. Con.,
58(2019), 585–601.
[2]
P. Yang, R. Ni, Z. Yao, et al., Robust contrast enhancement forensics using convolutional neural networks, (2018), arXiv preprint arXiv:1803.04749.
[3]
M. C. Stamm and K. J. R. Liu, Forensic estimation and reconstruction of contrast enhancement mapping, IEEE International Conference on Acoustics, Speech and Signal, (2010), 1698–1701.
[4]
G. Cao, Y. Zhao, R. Ni, et al., Contrast enhancement based forensics in digital images, IEEE T. Inf. Foren. Sec., 9(2014), 515–525.
[5]
X. Lin, C. T. Li and Y. Hu, Exposing image forgery through the detection of contrast enhancement, IEEE International Conference on Image Processing (ICIP), (2013), 4467–4471.
[6]
C. Yuan, X. Li, Q. M. Jonathan. Wu, et al., Fingerprint liveness detection from different fingerprint materials using convolutional neural network and principal component analysis, CMC-Comput. Mater. Con., 53(2017), 357–371.
[7]
F. Peng and D. L. Zhou, Discriminating natural images and computer generated graphics based on the impact of CFA interpolation on the correlation of PRNU, Digit. Invest., 11(2014), 111–119.
[8]
M. Long, F. Peng and Y. Zhu, Identifying natural images and computer generated graphics based on binary similarity measures of PRNU, Multimed. Tools. Appl., 78(2019), 489–506.
[9]
A. Radford, L. Metz and S. Chintala, Unsupervised representation learning with deep convolutional generative adversarial networks, (2015), arXiv preprint arXiv:1511.06434.
[10]
K. He, X. Zhang, S. Ren, et al., Deep residual learning for image recognition, IEEE International Conference on Computer Vision, (2016), 770–778.
[11]
Q. Cui, S. McIntosh and H. Sun, Identifying materials of photographic images and photorealistic computer generated graphics based on deep CNNs, CMC-Comput. Mater. Con., 55(2018), 229–241.
[12]
W. Quan, K. Wang, D. M. Yan, et al., Distinguishing between natural and computer-generated images using convolutional neural networks, IEEE T. Inf. Foren. Sec, 13(2018), 2772–2787.
[13]
T. Ng, S. Chang, J. Hs, et al., Columbia photographic images and photorealistic computer graphics dataset, ADVENT, Columbia University, (2005).
[14]
J. Wang, T. Li., X. Luo, et al., Identifying computer generated images based on quaternion central moments in color quaternion wavelet domain, IEEE T. Circ. Syst. Vid. Tec., (2018), DOI: 10.1109/TCSVT.2018.2867786.
[15]
G. Cao, Y. Zhao, R. Ni, et al., Anti-forensics of contrast enhancement in digital images, 12th ACM Workshop on Multimedia and Security, (2010), 25–34.
[16]
K. Singh, A. Kansal and G. Singh, An improved median filtering anti-forensics with better image quality and forensic undetectability, Multidi. Syst. Sign. P., (2019), 1–24.
[17]
A. Mehrish, A. V. Subramanyam and S. Emmanuel, Joint spatial and discrete cosine transform domain-based counter forensics for adaptive contrast enhancement. IEEE Access, 7(2019), 27183–27195.
[18]
P. M. Shelke and R. S. Prasad, An improved anti-forensics JPEG compression using least cuckoo search algorithm, Imaging. Sci. J., 66(2018), 169–183.
[19]
D. Kim, H. U. Jang, S. M. Mun, et al., Median filtered image restoration and anti-forensics using adversarial networks, IEEE Signal Proc. Let., 25(2018), 278–282.
[20]
M. C. Stamm and K. R. Liu, Anti-forensics of digital image compression, IEEE T. Inf. Foren. Sec., 6(2011), 1050–1065.
[21]
P. Yang, R. Ni, Y. Zhao, et al., Robust contrast enhancement forensics using convolutional neural networks, (2018), arXiv preprint arXiv:1803.04749.
[22]
Y. Luo, H. Zi, Q. Zhang, et al., Anti-forensics of jpeg compression using generative adversarial networks, 26th European Signal Processing Conference (EUSIPCO), (2018), 952–956.
[23]
H. Li, W. Luo, X. Qiu, et al., Identification of various image operations using residual-based features, IEEE T. Circ. Syst. Vid. Tec., 28(2018), 31–45.
[24]
R. Böhme and M. Kirchner, Counter-forensics: Attacking image forensics, Digital Image Forensics, Springer, New York, (2013), 327–366.
[25]
J. Fridrich and J. Kodovsky, Rich models for steganalysis of digital images, IEEE T. Inf. Foren. Sec., 7(2012), 868–882.
[26]
T. Pevny, P. Bas and J. Fridrich, Steganalysis by subtractive pixel adjacency matrix, IEEE T. Inf. Foren. Sec., 5(2010), 215–224.
[27]
I. Goodfellow, J. Pouget-Abadie, M. Mirza, et al., Generative adversarial nets, Advances in Neural Information Processing Systems, (2014), 2672–2680.
[28]
J. Hayes and G. Danezis, Generating steganographic images via adversarial training, Advances in Neural Information Processing Systems, (2017), 1954–1963.
[29]
D. Volkhonskiy, I. Nazarov, B. Borisenko, et al., Steganographic generative adversarial networks, (2017), arXiv preprint arXiv:1703.05502.
[30]
H. Shi, J. Dong, W. Wang, et al., SSGAN: Secure steganography based on generative adversarial networks, Pacific Rim Conference on Multimedia, Springer, Cham, (2017), 534–544.
[31]
R. Meng, S. G. Rice, J. Wang, et al., A fusion steganographic algorithm based on faster r-cnn, CMC-Comput. Mater. Con., 55(2018), 1–16.
[32]
S. Ren, K. He, R. Girshick, et al., Faster r-cnn: Towards real-time object detection with region proposal networks, Advances in Neural Information Processing Systems, (2015), 91–99.
[33]
I. Gulrajani, F. Ahmed, M. Arjovsky, et al., Improved training of wasserstein gans, Advances in Neural Information Processing Systems, (2017), 5767–5777.
[34]
G. Xu, H. Z. Wu, and Y. Q. Shi, Structural design of convolutional neural networks for steganalysis, IEEE Signal Proc. Let., 23(2016), 708–712.
[35]
L. Baroffio, L. Bondi, P. Bestagini, et al., Camera identification with deep convolutional networks. IEEE Signal Proc. Let., 24(2016), 259–263.
[36]
S. Xiang and H. Li, On the effect of batch normalization and weight normalization in generative adversarial networks, (2017), arXiv preprint arXiv:1704.03971.
[37]
Z. Liu, P. Luo, X. Wang, et al., Deep learning face attributes in the wild, IEEE International Conference on Computer Vision, (2015), 3730–3738.
[38]
T. Tieleman and G. Hinton, Lecture 6.5-rmsprop, coursera: neural networks for machine learning, University of Toronto, (2012).
Figure 1. The specific structure of the proposed architecture. The full name of the abbreviations NI and ICGI is natural images and improved computer graphic images. Each block represents a group of feature maps
Figure 2. The structure in detail of the residual block
Figure 3. The random sampled generated images by the well-trained model of the proposed approach
Figure 4. The loss curve of the test evolution during the iterations of 400,000
Figure 5. The loss curve of the discriminator on the real data during the training evolution
Figure 6. The loss curve of the discriminator on the fake data during the training evolution
Figure 7. The randomly selected target real natural images (left), and the generated images by the proposed approach targeting the randomly selected real natural images (right)
Figure 8. The loss curve graph of the discriminator during the test evolution
DownLoad: