Scribd
Upload
39 views14 pages

Deepfake Video Forensics Detection Tool Phase 1 Presentation Final

Uploaded by

itsnotme23.14
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
39 views14 pages

Deepfake Video Forensics Detection Tool Phase 1 Presentation Final

Uploaded by

itsnotme23.14
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 14

Deepfake Video Forensics

Detection Tool Using


Deep Learning
Group 11:

Aiswarya M- PTA21CC006
Anjali Manohar- PTA21CC013
Rasha K P- PTA21CC036
Sree Lakshmi S- PTA21CC049
Contents
01 02 03
Abstract Introduction Literature Review

04 05 06
Existing System Proposed System Design

07 08 09
Implementation Challenges and Risk Conclusion
Abstract
• Deep learning is important for detecting deepfakes, which are images and videos that
have been digitally manipulated.

• This project explores detecting deepfakes using Convolutional Neural Networks (CNNs)
and 3D CNNs, which are commonly used for both images and videos.

• These methods help identify fake media on social platforms, where deepfake creation tools
are increasingly available.

• This project examines current techniques, with CNNs focused on images and 3D CNNs
used for video analysis.

• By examining available datasets, challenges, and advanced methods, this work serves as a
resource for researchers and suggests future improvements for detecting deepfakes.
Introduction
• Deepfakes are digitally manipulated images and videos designed to imitate real content,
often making it challenging to distinguish between authentic and fabricated media.

• Detecting deepfakes has become essential in countering misinformation, particularly on


social media platforms where manipulated content can spread quickly and impact public
perception.

• Deep learning techniques, specifically Convolutional Neural Networks (CNNs) and 3D


CNNs, play a significant role in deepfake detection.

• While CNNs are effective for detecting manipulations in images, 3D CNNs handle temporal
sequences, making them suitable for video-based deepfake detection.

• This project focuses on reviewing current detection techniques, available datasets,


challenges, and state-of-the-art methods in the field.
Literature Review
• Deepfake Detection System: Combines models with MTCNN preprocessing,
improving detection accuracy in real-world scenarios.

• AVFF Audio-Visual Feature Fusion for Video Deepfake Detection: Uses both
audio and visual features to detect deepfakes.

• Deeepfake Detection with Combined Unsupervised-Supervised Contrastive


Learning: Combines unsupervised and supervised contrastive learning for improved
adaptability across diverse deepfake datasets.
Literature Review
• Deepfake Videos Generation and Detection: A Comprehensive Survey:
Reviews recent deepfake advancements and challenges, covering effectiveness across
various media formats.

• DeepSight: Enhancing deepfake Image Detection and Classification through


Ensemble and Deep learning Techniques: Introduce "DeepSight an ensemble
model for high-precision deepfake image detection.
Existing System
• CNN-Based Image Detection Models: CNNs are widely used to detect deepfake in
images by spotting features in single frames.

• RNN and LSTM Models for Temporal Analysis: These are used to catch
inconsistencies in videos by analyzing frame sequences.

• Hybrid Models (CNN + LSTM): These models combine CNN and LSTM to analyze
both spatial features in each frame and changes over time, making them more accurate
for video deepfakes.
Proposed System
• Video Frame Analysis: The tool analyzes multiple video frames at once to detect
unnatural movements.

• Audio-Video Synchronization Check: It checks if lip movements match the audio to


spot possible mismatches.

• Motion and Facial Feature Detection: The system examines head and facial
movements for anything that looks unrealistic.

• Artifact Detection: It identifies small visual errors like blurring that indicate
deepfake manipulation.

• User Interface and Reporting: An easy-to-use dashboard shows results clearly and
provides a detailed report.
Design
Implementation
• Data Collection and Preprocessing

• CNN Model for Image-Level Deepfake Detection

• 3D CNN Model for Video-Level Deepfake Detection

• Fusion Model to Combine CNN and 3D CNN Outputs

• Model Optimization and Testing

• Continuous Learning and Adaptation

• Deployment and Documentation


Challenges and Risk
• Data Requirements: Need for large, diverse, and high-quality video datasets.

• Privacy Issues: Handling sensitive content may violate privacy laws.

• High Maintenance: Ongoing updates and retraining needed; cost concerns.

• Ethical Misuse: Risk of tool being used for surveillance or false accusations.

• Dependency Risks: Vulnerability to outdated or insecure third-party libraries.


Conclusion
• The Deepfake Video Forensics Detection Tool addresses the pressing challenges posed
by deepfake technology in security and public trust.

• By employing advanced techniques like Convolutional Neural Networks (CNNs) and 3D


CNNs, we can effectively analyze both images and videos for manipulation.

• Our systematic approach includes comprehensive data collection, preprocessing, model


training, and optimization to ensure robust performance.

• The tool aims to significantly enhance media verification and forensic investigations,
providing reliable detection capabilities.
References
• "DeepFakes: A New Threat to Cybersecurity"T. S. D. Andrade, et al.IEEE Security &
Privacy, vol. 19, no. 6, pp. 54-62, Nov.-Dec. 2021.

• "Detecting Deepfake Videos Using Machine Learning: A Survey"R. C. Gupta, et al.IEEE


Access, vol. 9, pp. 147801-147820, 2021.

• "FakeCatcher: A DeepFake Detection Method Based on Motion Patterns"M. Yang, et


al.IEEE Transactions on Information Forensics and Security, vol. 16, pp. 1-15, 2021.

• "Towards Real-Time Detection of Deepfake Videos"Y. S. Choi, et al.2021 IEEE


International Conference on Image Processing (ICIP), pp. 1309-1313, 2021.

• "Video Forgery Detection Based on Deep Learning and Feature Fusion"Z. Liu, et
al.2021 IEEE International Conference on Multimedia and Expo (ICME), pp. 1-6, 2021.
Thank You!
CRÉDITOS: Este modelo foi criado pelo Slidesgo, e inclui ícones do
Slidesgo

Flaticon e infográficos e imagens da Freepik


Flaticon Freepik

You might also like