8. Deepfake Mix PPT using the CNN technique.pptx

Get simple presentation of deepfake detection

1. B y Abhishek Waghmare Talha Attar Tejas Rajput Guided by : DR. Rubeena Khan

2. 01 PROBLEM STATEMENT 03 04 FEASIBILITY AND SCOPE 02 OBJECTIVE LITERATURE REVIEW MOTIVATION

3. Problem Statement Develop a real-time deepfake detection system using CNN to distinguish between genuine and manipulated video content, addressing the growing threat of digital mis-information. The rise of deepfake technology has introduced significant challenges in maintaining the integrity of multimedia content on the internet and in various domains such as journalism, entertainment, and security.

4. MOTIVATION As increase in deepfakes it can be lead to tournish one’s reputation. Through deepfake one can trap an individual in a unauthourized activities. Criminals uses this technology to hide their original identity. Deepfake of leaders can lead to misguidance of people and degrade their personality

5. RESEARCH OBJECTIVE To identify such manipulations and distinguish them from real videos or images. Algorithms initially detect the whole face, or by using the face landmarks, they detect the face. Mask the identity of people's faces to protect their privacy. It also used in cyber crime whether the image, video is face swap or it had undergone a manipulation algorithm.

6. Paper Name : Purpose : Conclusion : Author : Publication/ Year : An Improved Dense CNN Architecture for Deepfake Image Detection. This paper presents a novel and improved deep-CNN (D-CNN) architecture for deepfake detection with reasonable accuracy and high generalizability. The proposed architecture offers 97.2% accuracy considering images from 5 different data sources for deepfake images and 2 different data sources for real images, Hence the proposed architecture provides a well-balanced performance over all data sources. YOGESH PATEL, SUDEEP TANWAR, PRONAYA BHATTACHARYA, RAJESH GUPTA1, AND TURKI ALSUWIAN IEEE-2023 Generalized Deepfake Video Detection Through Time-Distribution and Metric Learning Rapid advancements in the field of computer vision and AI have enabled the creation of synthesized images and videos known as deepfakes. Designed a generalized deepfake detector by creating a two-stream network that uses CNN-LSTM as its backbone. show that metric learning or contrastive loss function improves the overall effectiveness of a classification network. As the quality of deepfake images and videos constantly improves we believe that a more lasting solution would extract biological features and analyses face structures. Shahela Saif, Syed Sohaib Ali, Sumaira Kausar, and Amina Jameel IEEE-2022

7. Paper Name : Purpose : Conclusion : Author : Publication/ Year: MRE-Net: Multi-Rate Excitation Network for Deepfake Video Detection. This paper propose a novel Multi-Rate Excitation Network (MRE-Net) to effectively excite dynamic spatial-temporal inconsistency from the perspective of multiple rates for deepfake video detection. The proposed MRE-Net is composed of two components: Bipartite Group Sampling (BGS) and multiple rate branches. MIE captures the short-term and spatial divergence within groups, while LIE exploits the long-term temporal inconsistency between groups. The dynamic temporal inconsistency learned from multiple sampling rates further upgrades the generalization ability to unseen face-forged data. Guilin Pang , Baopeng Zhang , Zhu Teng , Member, IEEE, Zige Qi, and Jianping Fan IEEE-2023 An Exploratory Analysis on Visual Counterfeits Using Conv-LSTM Hybrid Architecture This work proposes a microscopic-typo comparison of video frames. This temporal-detection pipeline compares very minute visual traces on the faces of real and fake frames using Convolutional Neural Network (CNN) and stores the abnormal features for training. Visual forgery on videos and images can be automatically detected with precise accuracies from the proposed approach. As this architecture was built and trained on a humongous amount of data, any visual forgery can be easily detected giving the best accuracy. MOHAMMAD FARUKH HASHMI , B. KIRAN KUMAR ASHISH, AVINASH G. AND KESKAR3 IEEE-2020

8. Paper Name : Purpose : Conclusion : Author : Publication/ Year : Dynamic Difference Learning With Spatio–Temporal Correlation for Deepfake Video Detection Propose a novel dynamic difference learning method to distinguish between the inter-frame differences caused by face manipulation and the inter-frame differences caused by facial motions in order to model precise spatio-temporal inconsistency for deepfake video detection. DFDC-module and MSA-module are complementary and plug-and-play, they can provide any existing 2D CNNs with powerful spatio-temporal feature extraction capability, which can better cope with deepfake video detection. Qilin Yin , Wei Lu , Member, Bin Li , and Jiwu Huang IEEE-2023 Deepfake Detection on Social Media: Leveraging Deep Learning and FastText Embeddings for Identifying Machine-Generated Tweets Consequently, text-generative models have become increasingly powerful allowing the adversaries to use these remarkable abilities to boost social bots, allowing them to generate realistic deepfake posts and influence the discourse among the general public. Experimental results indicate that the design of the CNN architecture coupled with the utilization of FastText embeddings is suitable for efficient and effective classification of the tweet data with a superior 93% accuracy SAIMA SADIQ , TURKI ALJREES , AND SALEEM ULLAH1 IEEE-2023

9. FEASIBILITY The project would require about 16 GB of RAM and a stable internet connection which are easily accessible. Also, CNN which is an algorithm we plan to use in our project on it and is well documented, making it easy for us to study and implemented it.

10. Scope of Project The proposed CNN model can also be useful for problems like deepfake source identification. These problems become more challenging when powerful AI opensource toolkit and techniques like Generative Adversarial Networks (GANs) are improving its performance in creating perfect manipulation video . Hence our project aims to prevent AI from being misused and build trusted AI system and gain the attention of the research community.

11. By Abhishek Waghware Tejas Rajput Talha Attar Guided by: DR. Rubeena Khan

12. Data flow diagram 1 CNN and Yolo

13. Data flow diagram 2

14. DFD 3 CNN Yolo V3

15. ER Diagram

16. UML Diagram Class

17. ACTIVITY Feature Extraction Classification Using CNN and Yolo V3 algorithm No Overfitting

18. USE CASE Segmentation Feature Extraction

19. SEQUENCE

20. Software Requirement Specification RAM : 8 GB Hard Disk : 40 GB Processor : Intel i5 Processor  IDE : Spyder Coding Language : Python Version 3.8 Operating System : Windows 10 or 11

21. Hardware Requirement Specification Processor : Pentium-IV Speed : 1.1 GHz RAM :512 MB(min) Hard Disk : 40 GB