See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/382067030

Development of Deepfake Detection Techniques for Protecting Multimedia

Information using Deep Learning

Conference Paper · July 2024

CITATIONS READS

0 273

9 authors, including:

Jude Moses Anto Devakanth J Gowtham Raj

Madanapalle Institute of Technology & Science New Horizon College of Engineering
4 PUBLICATIONS 0 CITATIONS 3 PUBLICATIONS 0 CITATIONS

SEE PROFILE SEE PROFILE

R.Dhanapal Rd
Karpagam University
62 PUBLICATIONS 418 CITATIONS

SEE PROFILE

All content following this page was uploaded by R.Dhanapal Rd on 08 July 2024.

The user has requested enhancement of the downloaded file.

 Proceedings of the 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC-2024)
IEEE Xplore Part Number: CFP24BC3-ART; ISBN: 979-8-3503-7519-0

Development of Deepfake Detection Techniques

for Protecting Multimedia Information using
Deep Learning
2024 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC) | 979-8-3503-7519-0/24/$31.00 ©2024 IEEE | DOI: 10.1109/ICAAIC60222.2024.10575155

J Jude Moses Anto Devakanth Gowtham Raj

Siva Rama Lingham N
Department of Computer Science and Department of Computer Science and
Department of Computer Science and
Engineering Engineering
Engineering
Vel Tech Rangarajan Dr. Sagunthala Acharya Institute of Technology
Vel Tech Rangarajan Dr. Sagunthala
R&D Institute of Science and Bangalore, India
R&D Institute of Science and
Technology [email protected]
Technology
Chennai, India
Chennai, India
[email protected]
[email protected]
R.Janani Dr. R. Dhanapal
Gayathri K Department of Computer Science and
Department of Computer Science and
Department of Information Technology Engineering
Engineering
Bannari Amman Institute of Technolgy Karpagam Academy of Higher
Sri Krishna College of Engineering and
Erode, India Education,
Technology
[email protected] line 4: Coimbatore, India
Coimbatore, India
[email protected] [email protected]

Abstract—The development of proposed deepfake detection appearance and replicates them accordingly [2]. This is
techniques plays a prominent role in the safeguarding the originated as a tool for entertainment but rather formed as
multimedia information. The various challenges are addressed ethical concerns with potential misuse. The important
using the aid of deep learning techniques. This involves hybrid constraints deepfakes which involves the potential for various
optimization techniques such as particle swarm optimization
and genetic algorithms. This helps to improve the accuracy and
malicious attacks. This involves the creation of various
efficiency in the detection of deepfake. PSO helps in optimizing political speeches and images with misinformation
the weights and parameters of the neural network to obtain campaigns. Due to the advancement in technology, the role
faster convergence. GA helps in obtaining potential solutions to and creation of deepfakes are rising significantly [3]. Various
obtain the robust deepfake detection model. The integration of efforts are implemented to overcome the negative
deep learning with hybrid optimization involves the collection consequences of deepfakes through the development of
and preprocessing of the multimedia information and proceeds various deepfake detection tools. Various research and
to an optimization algorithm. The manipulated content in the developments are initiated to avoid the ill effects of these
online platform are detected during the training process. The deep fake technology with reducing the potential impact of
hybrid optimization tcehniques helps in obtaining model
generability and achieving resilience in various attacks
this technology. The manipulated content are detected using
occurring in the deepfake generation. The integration of PSO the robust tools and software.
and GA helps in the accurate detection of deepfake content
through continuous training and evaluation. Thus the proposed The role of artificial intelligence (AI) plays an
system forms a safeguarding tool for multimedia content and important role in the creation of deepfake technology. AI
helps to eliminate various risks. plays a central role in deepfake technology with the aid of
optimization algorithms. These algorithms helps in the
Keywords—Deepfake, Deep learning techniques, continuous amalgamation of the existing fabricated networks
Optimization algorithm, Genetic algorithm, Particle swarm
for the creation of the realistic content in the online platform.
optimization, Multimedia content
They become highly difficult and complicated to classify
between the dependable materials [4] The most important
I. INTRODUCTION aspect of AI in deepfake is the training phase.

Deepfake technology is defined as the integration of During the training process, the AI learns to
artificial intelligence (AI) and deep learning algorithms for replicate the exact copy of the user through analyzing the
the creation of realistic multimedia content and images. They behaviour and appearance. The training data involves various
are generally in the form of videos or audio recordings. The photos, videos and audio recordings. This helps in the
term "deepfake" is extracted from the combination of deep creation of an exact replica of the user. The development of
learning techniques. This is familiar due to the manipulation advanced technology leads to the rise of deepfakes. This
of human like facial gestures and hand gestures [1]. This is results in various negative consequences in the society. If the
highly complicated to identify deepfake images and videos quality of the deepfake may improve, it will become highly
through the naked eyes. The creation of deepfakes involves challenging to obtain the classification of real images and
the training of neural network layers on large amounts of content. This will become a challenging task both the humans
dataset which involves various images and videos. This and detection tools [5] If they are used for entertainment
algorithm learns the complete behaviour and physical purposes, they are highly invited. But due to certain negative

979-8-3503-7519-0/24/$31.00 ©2024 IEEE 786

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SRINAGAR. Downloaded on July 04,2024 at 11:10:05 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC-2024)
IEEE Xplore Part Number: CFP24BC3-ART; ISBN: 979-8-3503-7519-0

activities result in the depletion of the deepfake. This gives continuous learning model for accurate detection
rise to various security concerns. The false content spreading [9].
through the aid of deepfake technology results in losing the
trust in online platforms. They create a negative impact in the  The performance of the system are sensitive to
digital media. They also result in identity theft. The negative hyperparameters. Adam optimization must require
impact of deepfake techniques are avoided by using the aid proper tuning of learning rates and various
of powerful detection tools and mechanisms. This helps in parameters.
analyzing the information in multimedia for the detection and
identification of fake information [6]. There are various Thus the various constraints in the existing system are
regulations are put forward to overcome the the ethical and neglected using the aid of deep learning with hybrid
legal implications of deepfake technology. optimization techniques.

III. PROPOSED METHODOLOGY

The major objective of the proposed system is to

obtain deepfake detection system for overcoming various
constraints in multimedia. This is done using deep learning
techniques.

Multimedia data

Fig 1: Creation of deep fake images

Data augmentation with face
alignment
Figure 1 demonstrates the creation of deepfake images
through the aid of artificial intelligence.

Focus on facial features

II. EXISTING SYSTEM

The existing system in the detection and analysis of

the deepfake images are based on machine learning with Deepfake detection model
optimization techniques such as random forest algorithm, training using deep learning
decision tree and Adam optimization techniques. This results
in various drwabacks as listed below.
Model evaluation and
 Random forests and decision trees are prone to deployment
overfitting process. This occurs when the models are
trained on a limited dataset. This causes poor
generalization. This affects the performance on real- Fig 2: Stages in the proposed system
world deepfake content [7].
Figure 2 demonstrates the various stages in the proposed
 The extraction of relevant features from detecting system. The first stage involves the collection of data which
deepfakes are highly challenging task. The involves both the real and fake multimedia information.
effectiveness of the algorithm based on the quality These datasets are diverse in nature which involve varied
of the features. Manual feature engineering fails to background with lighting conditions. The obtained data is
capture the various intricate features in deepfake preprocessed. This involves image augmentation, facial
generation [8]. This results in suboptimal alignment and normalization process.
performance.
The high level features are extracted from the
 Training and deploying processes of random forests images and videos. They focus on the facial expression. Deep
and decision trees are a little bit costly. They are learning with optimization algorithms are used for the
susceptible to adversarial attacks and malicious training of deepfake detection model. This involves training
attacks. on both the real and fake datasets. Then the binary cross
entropy is analyzed and proceeded with optimization
 Random forests and decision trees lacks in techniques. The various performance matrices are obtained
interpretability compared to various models. The using the model evaluation techniques. Then the post-
development of deepfake technology needs a preprocessing techniques are initiated which involve
ensemble methods, confidence thresholding and verification

979-8-3503-7519-0/24/$31.00 ©2024 IEEE 787

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SRINAGAR. Downloaded on July 04,2024 at 11:10:05 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC-2024)
IEEE Xplore Part Number: CFP24BC3-ART; ISBN: 979-8-3503-7519-0

of temporal consistency. The developed system is integrated data. Advanced deep learning algorithms are used to
with various applications in the multimedia platforms. Real- overcome these threats [12]. These techniques helps in the
time detection is achieved for the detection of malicious automatic identification of malicious content in the
activities. After the implementation in the production multimedia platform. Feature extraction involves the
environment, continuous monitoring is important. This helps extraction and demonstration of the data from the input
in the prevention of various adversarial attacks. information. In the image and video analysis, feature
extraction involves the gathering of various facial patterns
and eye movements. Transfer learning helps in improving the
IV. DEEP LEARNING BASED DEEPFAKE DETECTION efficiency of the detection process. The frequent progression
SYSTEM of deepfake generation techniques provides iterative
refinement of the feature extraction process. The reliability of
The development of deepfake detection system using deep deepfake detection systems are obtained using the
learning techniques involves various stages. multimodal approaches [13]. This involves the combination
of both visual and auditory domains. This fusion of
Phase 1: Multimedia data collection information helps in obtaining information regarding the
multimedia content.
The development of deepfake detection techniques
involves the appropriate collection of multimedia data for Phase 2: Integration of deep learning with optimization
training the detection models. This helps in obtaining the algorithm
multimedia content with accurate precision. This dataset
involves various diverse information such as images, videos
and audio recordings. This helps to verify the versatility of
the detection models [10]. This helps in addressing the
various challenges in the online platform.

The collection of real world information requires

particular attention to get rid of various constraints because
the deepfake have the capability to replicate the real world
scenarios in online platforms. The dataset involves diverse
demographics and cultural backgrounds. The analysis of the
ethical acquisition of multimedia data is highly complicated.
They must include higher privacy in the collection of data
[11]. Anonymization techniques are used for the protection
of individual identity of the individual.

This helps in protecting the realistic nature of the

contents. The multimedia data collected should involve a
spectrum of deepfake creation techniques. This results in
various manipulation methods. After the collection of data,
the training and validating process takes place.

TABLE I: Multimedia data

Video ID File path Label

01 /path/to/video1 Genuine
02 /path/to/video1 Genuine
03 /path/to/video1 Genuine
04 /path/to/video1 Genuine
Fig 3: Deep learning in the detection of deepfake images
05 /path/to/video1 Genuine
Figure 3 represents the deep learning techniques which helps
in the detection of deepfake images.
Table I demonstrates the sample multimedia data obtained
from an online platform which involves video with its file Particle Swarm Optimization (PSO) is a powerful
path and label. tool for the development of deepfake detection techniques.
This helps to safeguard the multimedia information. The
The feature extraction plays a significant role in the realistic synthetic media is detected using deep learning
process of identifying the authentic content from manipulated techniques.
content in the online platform. Deepfakes result in a
significant threat to the trust and intedgrity of the multimedia

979-8-3503-7519-0/24/$31.00 ©2024 IEEE 788

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SRINAGAR. Downloaded on July 04,2024 at 11:10:05 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC-2024)
IEEE Xplore Part Number: CFP24BC3-ART; ISBN: 979-8-3503-7519-0

The integration of PSO in the development of through integrating the strengths of the various detection
deepfake detection have the ability to optimize the various algorithms. Figure 5 demonstrates the fitness function.
parameters of the complex models. PSO is inspired by the
process of natural selection which helps in effectively fine-
tune the myriad parameters. The important advantages of the Phase 3: Ensemble technology
PSO optimization technique involves identification of
optimal configurations that helps in the deepfake content Ensemble technology is an important parameter in
more accurately [14]. This helps in obtaining robust and the development of robust deepfake detection techniques.
reliable configurations. They tend to provide a deviation from Ensemble techniques includes the accumulation of diverse
manipulated content and original content through iterations. deepfake detection models. They helps in the identification
of manipulated content in diverse phenomena. This helps the
The adaptability of deepfake detection techniques is overall system to obtain various features and patterns
achieved using PSO. This is achieved through continuous associated with the deepfake generation [18]. This helps in
learning and evolution. They also helps in obtaining dynamic obtaining a resilient system against manipulation techniques.
alterations to the recognition models [15]. The PSO helps in The ensemble technology involves the fusion of various
the identification of various patterns with the deepfake technology in the detection process which involves the image
methods. and audio analysis. This multidimensional approach helps in
increasing the accuracy of the deepfake detection system.
They also provide continuous learning and adaptation in the
diverse platform for the originating deepfake techniques.
They increase the system's reliability through adopting the
level of redundancy. They also helps in the reduction of false
positives or negatives [19].

Fig 4: Detection of deepfakes

The important challenges in the effective Fig 5: Fitness function

detection of deepfake involves the complexity and non-
linearity of the patterns and features in the multimedia. The Phase 4: Threshold optimization
genetic algorithm is used to handle the complex optimization
problems through the natural selection process [16]. They Threshold optimization involves the detection of the
helps in improving the successive generations. They are used delicate balance that increases the system's true positive rate
for optimize the parameters and helps in the identification of that denotes actual deepfake with reducing the false positive
the various manipulations in multimedia content as shown in rate which denotes the original content as deepfakes.
figure 4. They also helps in the identification of the Obtaining proper equilibrium is important for a practical and
discriminative features. reliable deepfake detection system. There are various factors
responsible for the complexity of threshold optimization in
GA tend to improve the system accuracy in the deepfake detection. During the training process, the threshold
detection of the detection systems. The fine tuning of the value is modified or adjusted based on the model
performance parameter are achieved through the performance on the training set. This involves temporal
optimization of hyperparameters for obtaining deepfake analysis with modelling of their behaviour. Continuous
detection models. The dynamic nature of the deepfake model collaboration is highly important for the identification of
involves the continuous adaptation of detection techniques various threats in deepfake technology [20]. Then the real
[17][21]. They helps in obtaining mechanisms for dynamic time implementation takes place. This includes the
model evolution. integration of hardware systems. This involves the graphics
processing units (GPUs).
The development of ensemble models are
implemented using a genetic algorithm. This is achieved

979-8-3503-7519-0/24/$31.00 ©2024 IEEE 789

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SRINAGAR. Downloaded on July 04,2024 at 11:10:05 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC-2024)
IEEE Xplore Part Number: CFP24BC3-ART; ISBN: 979-8-3503-7519-0

V. EXPERIMENTAL DISCUSSION AND ANALYSIS 400

90
350
The proposed system is evaluated in the proteus
300 76
platform. They provides overall performance of the 92
functioning of hybrid optimization techniques through 250
various performance metrics. 85
200 96
TABLE II: Result analysis 150 80
100 99
Video File path Label Manipulation
70
50
ID
0 0 0
01 /path/to/video1 Genuine Face Proposed Existing
-50
Replacement system system

02 /path/to/video1 Genuine Voice Accuracy Precision

Replacement Recall Energy efficiency

03 /path/to/video1 Genuine Lip sync

Fig 7: Performance metrics
04 /path/to/video1 Genuine None
05 /path/to/video1 Genuine None The performance metrics of the proposed hybrid optimization
techniques are shown in figure 7.
Table II demonstrates the result analysis which identifies the These hardware system helps in the processing and
manipulation of data using deep learning techniques. analysis of multimedia data. They provide various
possibilities for the detection of deepfakes in real-time
environments. The scalability of the system is achieved using
the parallel computing techniques and distributed computing.
This helps in handling large amounts of multimedia data. The
integration of the AI system helps in obtaining transparency
in the detection system. They are highly important for
sensitive applications. The continuous updates and model
retraining are essential for the ever-changing condition of the
deepfake techniques. An adaptive system helps to adopt
innovative variations for the deepfakes against various
malicious attacks.

The various performance metrics such as accuracy,

precision, recall and F1 score are calculated to determine the
functioning of the system.

Fig 6: ROC curve VI. CONCLUSION

Figure 6 demonstrates the ROC curve. This helps to analyze

The development of deepfake detection techniques
the performance of a binary classification proposed model
using deep learning provides a significant tool for protecting
across the various threshold values. They represent the trade-
the multimedia information. The evolution of deepfake
off between the true positive rate and the false positive rate.
technology results in various challenges to the authenticity
and integrity of digital information in online platforms. The
proposed system helps in the creation of robust detection
mechanisms. This helps in increasing the resilience of
multimedia information against deepfakes. The integration of
deep learning techniques with optimization algorithms helps
in the development of efficient deepfake detection systems.
The continuous refinement of the proposed techniques
integrated with large-scale datasets with feature extraction
helps in improving the overall performance of the system.
This helps in improving the reliability of deepfake detection
systems. Thus the proposed system helps to maintain trust
and authenticity in multimedia platforms.

979-8-3503-7519-0/24/$31.00 ©2024 IEEE 790

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SRINAGAR. Downloaded on July 04,2024 at 11:10:05 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC-2024)
IEEE Xplore Part Number: CFP24BC3-ART; ISBN: 979-8-3503-7519-0

International Journal for Research in Applied Science and Engineering

Technology, 11(3). https://doi.org/10.22214/ijraset.2023.49681
REFERENCES [14] Salman, S., Shamsi, J. A., & Qureshi, R. (2023). Deep Fake Generation
and Detection: Issues, Challenges, and Solutions. IT Professional,
25(1). https://doi.org/10.1109/MITP.2022.3230353
[1] Liao, X., Wang, Y., Wang, T., Hu, J., & Wu, X. (2023). FAMM: Facial [15] Abdullah, M. T., & Ali, N. H. M. (2023). DeepFake Detection
Muscle Motions for Detecting Compressed Deepfake Videos over Improvement for Images Based on a Proposed Method for Local
Social Networks. IEEE Transactions on Circuits and Systems for Video Binary Pattern of the Multiple-Channel Color Space. International
Technology. https://doi.org/10.1109/TCSVT.2023.3278310 Journal of Intelligent Engineering and Systems, 16(3).
[2] Mehra, A., Agarwal, A., Vatsa, M., & Singh, R. (2023). Motion https://doi.org/10.22266/ijies2023.0630.07
Magnified 3-D Residual-in-Dense Network for DeepFake Detection. [16] Yang, W., Zhou, X., Chen, Z., Guo, B., Ba, Z., Xia, Z., Cao, X., & Ren,
IEEE Transactions on Biometrics, Behavior, and Identity Science, 5(1). K. (2023). AVoiD-DF: Audio-Visual Joint Learning for Detecting
https://doi.org/10.1109/TBIOM.2022.3201887 Deepfake. IEEE Transactions on Information Forensics and Security,
[3] Lin, H., Huang, W., Luo, W., & Lu, W. (2023). DeepFake detection 18. https://doi.org/10.1109/TIFS.2023.3262148
with multi-scale convolution and vision transformer. Digital Signal [17] Kingra, S., Aggarwal, N., & Kaur, N. (2023). SiamNet: Exploiting
Processing: A Review Journal, 134. source camera noise discrepancies using Siamese Network for
https://doi.org/10.1016/j.dsp.2022.103895 Deepfake Detection. Information Sciences, 645.
[4] Salvi, D., Liu, H., Mandelli, S., Bestagini, P., Zhou, W., Zhang, W., & https://doi.org/10.1016/j.ins.2023.119341
Tubaro, S. (2023). A Robust Approach to Multimodal Deepfake [18] Wang, T., Cheng, H., Chow, K. P., & Nie, L. (2023). Deep
Detection. Journal of Imaging, 9(6). Convolutional Pooling Transformer for Deepfake Detection. ACM
https://doi.org/10.3390/jimaging9060122 Transactions on Multimedia Computing, Communications and
[5] Usmani, S., Kumar, S., & Sadhya, D. (2023). Efficient deepfake Applications, 19(6). https://doi.org/10.1145/3588574
detection using shallow vision transformer. Multimedia Tools and [19] Matukumalli, V., Naga Sasidhar Maddi, S., Krishna Angirekula, K.,
Applications. https://doi.org/10.1007/s11042-023-15910-z Reddy Pulicherla, V., Senthil kumar, A. M., Maridurai, T., …
[6] Coccomini, D. A., Caldelli, R., Falchi, F., & Gennaro, C. (2023). On Kasinathan, D. (2021). Augment reality chatbot using cloud. Materials
the Generalization of Deep Learning Models in Video Deepfake Today: Proceedings, 46, 4254–4257.
Detection. Journal of Imaging, 9(5). doi:10.1016/j.matpr.2021.03.058.
https://doi.org/10.3390/jimaging9050089 [20] Enireddy, V., Finney Daniel shadrach, S., Shobha rani, P., Anitha, R.,
[7] Xu, Y., Raja, K., Verdoliva, L., & Pedersen, M. (2023). Learning Vallinayagam, S., Maridurai, T., … Balakrishnan, E. (2021).
Pairwise Interaction for Generalizable DeepFake Detection. Prediction of human diseases using optimized clustering techniques.
Proceedings - 2023 IEEE/CVF Winter Conference on Applications of Materials Today: Proceedings, 46, 4258–4264.
Computer Vision Workshops, WACVW 2023. doi:10.1016/j.matpr.2021.03.068.
https://doi.org/10.1109/WACVW58289.2023.00074 [21] Ragupathy, P., & Vivekanandan, P. (2019). Towards Enhanced
[8] Janutėnas, L., Janutėnaitė-Bogdanienė, J., & Šešok, D. (2023). Deep Anomaly Object Detection and Face Recognition (EAODFR) in
Learning Methods to Detect Image Falsification. Applied Sciences Surveillance Videos using Recurrent Neural Networks. International
(Switzerland), 13(13). https://doi.org/10.3390/app13137694 Journal of Recent Technology and Engineering (IJRTE), 8(4), 3597-
[9] Chen, H., Li, Y., Lin, D., Li, B., & Wu, J. (2023). Watching the BiG 3603. doi:10.35940/ijrte.D7828.118419
artifacts: Exposing DeepFake videos via Bi-granularity artifacts.
Pattern Recognition, 135.
https://doi.org/10.1016/j.patcog.2022.109179
[10] Liang, P., Liu, G., Xiong, Z., Fan, H., Zhu, H., & Zhang, X. (2023). A
facial geometry based detection model for face manipulation using
CNN-LSTM architecture. Information Sciences, 633.
https://doi.org/10.1016/j.ins.2023.03.079
[11] Han, R., Wang, X., Bai, N., Wang, Q., Liu, Z., & Xue, J. (2023). FCD-
Net: Learning to Detect Multiple Types of Homologous Deepfake Face
Images. IEEE Transactions on Information Forensics and Security, 18.
https://doi.org/10.1109/TIFS.2023.3269152
[12] Patel, Y., Tanwar, S., Bhattacharya, P., Gupta, R., Alsuwian, T.,
Davidson, I. E., & Mazibuko, T. F. (2023). An Improved Dense CNN
Architecture for Deepfake Image Detection. IEEE Access, 11.
https://doi.org/10.1109/ACCESS.2023.3251417
[13] Gadgilwar, J., Rahangdale, K., Jaiswal, O., Asare, P., Adekar, P., &
Bitla, Prof. L. (2023). Exploring Deepfakes - Creation Techniques,
Detection Strategies, and Emerging Challenges: A Survey.

979-8-3503-7519-0/24/$31.00 ©2024 IEEE 791

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SRINAGAR. Downloaded on July 04,2024 at 11:10:05 UTC from IEEE Xplore. Restrictions apply.

View publication stats