Deepfake Detection_Using Machine Learning .pptx

Deep fake Detection

1. Deepfake Detection u sing ResNext & LSTM Under the guidance of Dr. Ajaya Kumar Dash Satyaprakash Sahoo ID: B220049 Akash Parida ID: B420005 Bharat Patidar ID: B420014

2. WHAT ARE DEEPFAKES ? Deepfakes are synthetic media crafted using sophisticated deep learning techniques like GANs and autoencoders. These algorithms learn from large datasets of images and videos to create realistic content.

3. Literature Survey Paper Year Author Model FaceForensics++: Learning to Detect Manipulated Facial Images 2019 A Rossler et a l . XceptionNet-CNN A Novel Machine Learning based Method for Deepfake Video Detection in Social Media 2020 A Mitra et al. XceptionNet-CNN iCaps-Dfake: An Integrated C apsule-Based Model for Deepfake Image and Video Detection 2021 SS Khalil et al. HRNet-CNN A machine mearning based approach for deepfake detection in social media through key video frame extraction 2021 A Mitra et al. ResNet-50, Xception

4. Problem Statement Training a Deepfake Detection model on the available data sets ; t hat can provide impressive performance score

5. Datasets FaceForensics++ DFDC 1000 original video sequences that have been manipulated with four automated face manipulation methods Deepfake Detection Challenge (DFDC) dataset consisting of 5K videos featuring two facial modification algorithms Large-scale challenging dataset for deepfake forensics. It includes 590 original videos and 5639 corresponding DeepFake videos Celeb-DF

6. Proposed Workflow

7. LSTM Models Integrates gated mechanisms for processing sequential data and capturing long-term dependencies Processes feature representations from ResNeXt-50, capturing temporal dynamics in video data. F or Feature Extraction ResNeXt-50 Addresses gradient issues with skip connections, enhancing trainability. Cardinality in ResNeXt-50 enables multi-path ensemble for improved feature representation F or Temporal Modeling

8. Model Summary

9. Evaluation Metrics Accuracy The fraction of predictions the model got right Precision How often the model correctly identifies positive cases Recall The model’s ability to capture all positive cases F1 Score Balanced measure of the model’s performance, considering both false positives and false negatives

10. Result and Discussion %

11. Confusion matrix

12. Obtained Results Accuracy 91.5% 96.1% 91.4% 93.8% Recall Precision F1 Score

13. Conclusion In this project work, the proposed ResNeXt-LSTM model for Deepfake Detection for the combined dataset demonstrates promising accuracy and robust performance.

14. Future Scope Integration with real-time processing and edge computing is crucial for deploying deepfake detection in real-world settings. Continuous dataset augmentation and diverse model retraining are essential for keeping up with evolving deepfake techniques.

15. Thank you !