(7KS09) Seminar Presentation

on

Deep Learning for DeepFakes: Creation,

Detection and Impact

Department of Computer Science & Engineering

Prof. Ram Meghe Institute of Technology & Research, Badnera - Amravati
2022-2023
Content
• Abstract
• Introduction
• Literature Survey
• Types of DeepFake with its applications
• DeepFake vs Shallowfake
• DeepFake: Creation
• DeepFake: Detection
• DeepFake: Prevention
• Conclusion and Future Scope
• References
 Abstract
• “DeepFake” is derived from “Deep Learning” and “Fake”.
• It offers automated procedures to create fake content that is
harder for human observers to detect.
• It can cause threats to privacy, democracy and national
security.
• Therefore, proposal of technologies that can automatically
detect and assess the integrity of digital media is important.
 Introduction
• Image or recording that has been altered to misrepresent
someone as doing something that was not actually done.
• For example, FaceApp
• It is AI enabled software tool used for realistic-looking face
swapping in images and videos.
• To make a realistic deepfake, it requires large amount of data
to train models.
• It facilitates numerous possibilities in various domain like
education, art etc.
 Literature Survey

This section outlines what has been done so far in Deep Learning for DeepFakes: Creation, Detection
and Impact.
McCloskey et al. [24] presented an approach to identify fake images by employing the fact that the
color information is evidently dissimilar between real camera and fake synthesis samples. The color
key-points from input samples were used to train the SVM for classification. This approach exhibits
better fake sample detection accuracy, however, it may not perform well for blurred images.
Y. Li et al. [27] proposed a technique to detect deepfakes by using the fact that the manipulated videos
lack accurate eye blinking in synthesized faces. CNN approach was used to detect the lack of eye
blinking in the videos to expose the forged content. This technique shows better deepfake detection
performance, however, it only uses the lack of eye blinking as a clue to detect the deepfakes.
This approach has the following potential limitations:
i) It is unable to detect the forgeries in videos with frequent eye blinking.
ii) It is unable to detect manipulated faces with closed eyes in training.
iii) It is inapplicable in scenarios where forgers can create realistic eye blinking in synthesized faces.
Types of DeepFakes with their Applications
 DeepFake vs Shallowfake
Differences:
• Method of production
• Involvement of deep learning
• AI generated

Similarities:
• Goal is to twist reality
• Utilization of pre-existing media content
• Tool of misinformation campaign
• Both should be utilized with caution
 DeepFake: Creation
• Creator first train a neural network on many hours of real video
footage of the person to give it a realistic understanding of what the it
looks from different angles and under different lighting.

• Then they’d combine the trained network with computer-graphics

techniques to superimpose a copy of the person onto a different actor.

• A class of deeplearning algorithms known as GANs [Generative

Adversarial Network] is used in deepfake development.
 Generative Adversarial Networks [GANs]

• Generative modeling is an unsupervised learning task automatically discovers patterns

in input data in such a way that the model can be used to generate new examples .

• GANs are a clever way of training a generative model by framing the problem as a
supervised learning problem with two sub-models:

• The generator model that we train to generate new examples, and the discriminator
model that tries to classify examples as either real (from the domain) or fake
(generated).

• The two models are trained together in a zero-sum game, adversarial, until the
discriminator model is fooled about half the time, meaning the generator model is
generating plausible examples.
Figure: The GAN architecture consisting of a generator and a discriminator, and each can be
implemented by a neural network. The entire system can be trained with backpropagation that
allows both networks to improve their capabilities.
 DeepFake: Detection

Figure: Example of characteristics that may indicate a deepfake.

 DeepFake: Prevention

Figure[3] Preventive measures against DeepFake

 Conclusion and Future Scope

Conclusion:
• Deep learning can be used in deepfake creation and detection methods.

• Deepfakes are created using generative adversarial networks [GANs], in which

two machine learning models exists. One model trains on a dataset and the other
model tries to detect the deepfakes.

• Deepfakes is also being used to create fake news, videos, images, and terrorism
events that can cause social, and financial fraud.
Future Scope:

• Future work will focus on evaluating different detection methods by using real
and manipulated datasets.

• The advancement in DeepFake technology might improve fullbody deepfakes.

But, it will be difficult to detect by the existing detection techniques.

• Deepfakes will continue to evolve and spread further, and issues like the lack of
details in the synthesis will be overcome.

• GANs will be the main drivers of deepfakes development in the future.

References
[1] A. Malik, M. Kuribayashi, S. M. Abdullahi and A. N. Khan, ”DeepFake Detection for Human Face Images and
Videos: A Survey,” in IEEE Access, vol. 10, pp. 18757-18775, 2022, doi: 10.1109/ACCESS.2022.3151186.
[2] N. Waqas, S. I. Safie, K. A. Kadir, S. Khan and M. H. Kaka Khel, ”DEEPFAKE Image Synthesis for Data
Augmentation,” in IEEE Access, vol. 10, pp. 80847- 80857, 2022, doi: 10.1109/ACCESS.2022.3193668.
[3] M. S. Rana, M. N. Nobi, B. Murali and A. H. Sung, ”Deepfake Detection: A Systematic Literature Review,” in IEEE
Access, vol. 10, pp. 25494-25513, 2022, doi: 10.1109/ACCESS.2022.3154404.
[4] A. H. Khalifa, N. A. Zaher, A. S. Abdallah and M. W. Fakhr, ”Convolutional Neural Network Based on Diverse
Gabor Filters for Deepfake Recognition,” in IEEE Access, vol. 10, pp. 22678-22686, 2022, doi:
10.1109/ACCESS.2022.3152029.
[5] Y. Li et al., ”DeepFake-o-meter: An Open Platform for DeepFake Detection,” 2021 IEEE Security and Privacy
Workshops (SPW), 2021, pp. 277-281, doi: 10.1109/SPW53761.2021.00047.
[6] L. Guarnera, O. Giudice and S. Battiato, ”Fighting Deepfake by Exposing the Convolutional Traces on Images,” in
IEEE Access, vol. 8, pp. 165085-165098, 2020, doi: 10.1109/ACCESS.2020.3023037.
[7] Thanh Thi Nguyena, Quoc Viet Hung Nguyenb and Dung Tien Nguyena, ”Deep Learning for Deepfakes Creation
and Detection: A Survey” 2022, arXiv:1909.11573v5
[8] S. Reed, Z. Akata, X. Yan, L. Logeswaran, B. Schiele, and H. Lee, “Generative adversarial text to image
synthesis,” 2016, arXiv:1605.05396. [Online]. Available: http://arxiv.org/abs/1605.05396
[9] M.-H. Maras and A. Alexandrou, “Determining authenticity of video evidence in the age of artificial intelligence
and in the wake of deepfake videos,” Int. J. Evidence Proof, vol. 23, no. 3, pp. 255–262, Jul. 2019, doi:
10.1177/1365712718807226
[10] I. Goodfellow, J. P. Abadie, M. Mirza, B. Xu, D. W. Farley, S. Ozair, A. Courville, and Y. Bengio, “Generative
adversarial nets,” in Proc. 27th Int. Conf. Neural Inf. Process. Syst. (NIPS), vol. 2. Cambridge, MA, USA: MIT
Press, 2014, pp. 2672–2680
[11] G. Patrini, F. Cavalli, and H. Ajder, “The state of deepfakes: Reality under attack,” Deeptrace B.V.,
Amsterdam, The Netherlands, Annu. Rep. v.2.3., 2018.
[12] F. F. Kharbat, T. Elamsy, A. Mahmoud and R. Abdullah, ”Image Feature Detectors for Deepfake Video
Detection,” 2019 IEEE/ACS 16th International Conference on Computer Systems and Applications (AICCSA),
2019, pp. 1-4, doi: 10.1109/AICCSA47632.2019.9035360.
[13] N. Kumar, P. P, V. Nirney and G. V, ”Deepfake Image Detection using CNNs and Transfer Learning,” 2021
International Conference on Computing, Communication and Green Engineering (CCGE), 2021, pp. 1-6, doi:
10.1109/CCGE50943.2021.9776410.
[14] K. Lin, W. Han, Z. Gu and S. Li, ”A Survey of DeepFakes Generation and Detection,” 2021 IEEE Sixth
International Conference on Data Science in Cyberspace (DSC), 2021, pp. 474-478, doi:
10.1109/DSC53577.2021.00076.
[15] A. M. Saber et al., ”DeepFake Video Detection,” 2022 2nd International Mobile, Intelligent, and Ubiquitous
Computing Conference (MIUCC), 2022, pp. 425-431, doi: 10.1109/MIUCC55081.2022.9781791.
[16] S. P and S. Sk, ”DeepFake Creation and Detection:A Survey,” 2021 Third International Conference on
Inventive Research in Computing Applications (ICIRCA), 2021, pp. 584-588, doi:
10.1109/ICIRCA51532.2021.9544522.
[17] S. A. Aduwala, M. Arigala, S. Desai, H. J. Quan and M. Eirinaki, ”Deepfake Detection using GAN
Discriminators,” 2021 IEEE Seventh International Conference on Big Data Computing Service and Applications
(BigDataService), 2021, pp. 69-77, doi: 10.1109/BigDataService52369.2021.00014.
[18] A. Malik, M. Kuribayashi, S. M. Abdullahi and A. N. Khan, ”DeepFake Detection for Human Face Images and
Videos: A Survey,” in IEEE Access, vol. 10, pp. 18757-18775, 2022, doi: 10.1109/ACCESS.2022.3151186.
[19] C. Yang, L. Ding, Y. Chen and H. Li, ”Defending against GAN-based DeepFake Attacks via Transformation-
aware Adversarial Faces,” 2021 International Joint Conference on Neural Networks (IJCNN), 2021, pp. 1-8, doi:
10.1109/IJCNN52387.2021.9533868.
[20] S. Ramachandran, A. V. Nadimpalli and A. Rattani, ”An Experimental Evaluation on Deepfake Detection
using Deep Face Recognition,” 2021 International Carnahan Conference on Security Technology (ICCST), 2021,
pp. 1-6, doi: 10.1109/ICCST49569.2021.9717407.
[21] N. Waqas, S. I. Safie, K. A. Kadir, S. Khan and M. H. Kaka Khel, ”DEEPFAKE Image Synthesis for Data
Augmentation,” in IEEE Access, vol. 10, pp. 80847- 80857, 2022, doi: 10.1109/ACCESS.2022.3193668. 27 Deep
Learning for DeepFakes: Creation, Detection and Impact
[22] T. Jung, S. Kim and K. Kim, ”DeepVision: Deepfakes Detection Using Human Eye Blinking Pattern,” in IEEE
Access, vol. 8, pp. 83144-83154, 2020, doi: 10.1109/ACCESS.2020.2988660
[23] A. S. U¸can, F. M. Bu¸cak, M. A. H. Tutuk, H. ˙I. Aydin, E. Semiz and S¸. Bahtiyar, ”Deepfake and Security of
Video Conferences,” 2021 6th International Conference on Computer Science and Engineering (UBMK), 2021,
pp. 36-41, doi: 10.1109/UBMK52708.2021.9558963.
[24] S. McCloskey and M. Albright, ”Detecting gan-generated imagery using color cues,” arXiv preprint 2018,
arXiv:08247