Systems and Means of Informatics

2023, Volume 33, Issue 2, pp 25-33

DEEPFAKE IMAGE DETECTION USING BISPECTRAL ANALYSIS

S. P. Nikitenkova

Abstract

As deep-fake image synthesis tools become more powerful and available, there is a growing need to develop methods for detecting generated content. The main goal of the work is to test the application of bispectral analysis as a tool for detecting images generated by artificial intelligence (AI).
It is shown that higher-order spectral correlations detected by spectral analysis are less present in natural images compared to the images generated using generative-adversarial neural networks GAN (generative adversarial network).
These correlations are probably the result of fundamental properties of the image generation process. The clustering procedure has shown encouraging results: it determines the generated images with an accuracy of 80%.

[+] References (21)

Nightingale, S., and H. Farid. 2022. Synthetic faces are more trustworthy than real faces. J. Vision 22(14):3068. doi: 10.1167/jov.22.14.3068.
Goodfellow, I., J. Pouget-Abadie, M. Mirza, et al. 2020. Generative adversarial networks. Commun. ACM 63(11): 139-144. doi: 10.1145/3422622.
Matern, F., C. Riess, and M. Stamminger. 2019. Exploiting visual artifacts to expose deepfakes and face manipulations. Winter Applications of Computer Vision Workshops. IEEE. 83-92. doi: 10.1109/WACVW.2019.00020.
Pavelyeva, E. A. 2013. Poisk sootvetstviy mezhdu klyuchevymi tochkami izobrazheniy raduzhnykh obolochek glaz s pomoshch'yu metoda proektsionnoy fazovoy korrelyatsii [The search for matches between the iris key points using Hermite projection phase- only correlation method]. Sistemy i Sredstva Informatiki - Systems and Means of Informatics 23(2):74-88. doi: 10.14357/08696527130206.
Karras, T., S. Laine, M. Aittala, J. Hellsten, J. Lehtinen, and T. Aila. 2020. Analyzing and improving the image quality of StyleGAN. IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE. 8110-8119.
Durall, R., M. Keuper, F.-J. Pfreundt, and J. Keuper. 2019. Unmasking deepfakes with simple features. arXiv.org. 8 p. Available at: https://arxiv.org/abs/1911.00686 (accessed April 17, 2023).
Frank, J., T. Eisenhofer, L. Schonherr, A. Fischer, D. Kolossa, and T. Holz. 2020. Leveraging frequency analysis for deep fake image recognition. 37th Conference (International) on Machine Learning Proceedings. 119:3247-3258.
Dong, С., A. Kumar, and E. Liu. 2022. Think twice before detecting GAN-generated fake images from their spectral domain imprints. IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE. 7865-7874.
Ng, T.T., S.F. Chang, and Q. Sun. 2004. Blind detection of photomontage using higher order statistics. Symposium (International) on Circuits and Systems. IEEE. 5:1329901. doi: 10.1109/ISCAS.2004.1329901.
Wen, Z., D. Fraser, and A. Lambert. 2009. Bicoherence: Anew lucky region technique in anisoplanatic image restoration. Appl. Optics 48(32):6111-6119.
Hinich, M. J., and M. A. Wolinsky. 1988. A test for aliasing using bispectral analysis. J. Am. Stat. Assoc. 83(402):499-502.doi: 10.1080/01621459.1988.10478623.
Mendel, J. M. 1991. Tutorial on higher-order statistics (spectra) in signal processing and system theory: Theoretical results and some applications. P. IEEE 79(3):278-305. doi: 10.1109/5.75086.
NVlabs/FFHQ-dataset. Available at: https://github.com/NVlabs/ffhq-dataset (accessed April 17, 2023).
NVlabs/StyleGAN2-dataset. Available at: https://github.com/NVlabs/stylegan2 (accessed April 17, 2023).
High-resolution anime face dataset (512x512). Available at: https://www.kaggle.com/ datasets/subinium/highresolution-anime-face-dataset-512x512 (accessed April 17, 2023).
GitHub - justinpinkney/awesome-pretrained-StyleGAN2-Anime-portraits-dataset. Available at: https: //github.com/justinpinkney/awesome-pretrained-stylegan2# Anime-portraits (accessed April 17, 2023).
Silvagni, D., С. Djerroud, T. Reveille, and E. Gravier. 2018. Bispectral analysis: Comparison of two windowing functions. J. Phys. Conf. Ser. 982(1):012007. doi: 10.1088/1742-6596/982/1/012007.
Scikit-learn: Machine learning in Python. Available at: http://scikit-learn.org (accessed April 17, 2023).
Ng, A. 2018. Machine learning yearning: Technical strategy for AI engineers, in the era of deep learning. Self-publishing. 118 p.
Karras, T., M. Aittala, S. Laine, et al. 2021. Alias-free generative adversarial networks. Adv. Neur. In. 34:852-863.
Alaluf, Y., O. Patashnik, Z. Wu, A. Zamir, E. Shechtman, D. Lischinski, and D. Cohen-Or. 2023. Third time's the charm? Image and video editing with StyleGAN3. Computer vision. Eds. L. Karlinsky, T. Michaeli, and K. Nishino. Lecture notes in computer science ser. Cham: Springer. 13802:204-220. doi: 10.1007/978-3-031- 25063-7_13.

[+] About this article