Scribd
Upload
18 views11 pages

Group 4

Uploaded by

zade.ameyy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
18 views11 pages

Group 4

Uploaded by

zade.ameyy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 11

RAJIV GANDHI COLLEGE OF ENGINEERING, RESEARCH & TECHNOLOGY, CHANDRAPUR

(DR. BABASAHEB AMBEDKAR TECHNOLOGICAL UNIVERSITY, LONERE)


(2024-25)

Project Seminar
Group no.: 4

“Deepfake
Detection”
Amey Zade-CSEA705
Kaushal Kamde-CSEA741
Sanika Gongale-CSEA737
Dr. Nitin Janwe Aakanksha Toutam-CSEA701 Dr. Manisha Pise
HOD, CSE Vidish Worah-CSEA769 Project Incharge
2

CONTENTS
• Abstract
• Introduction
• Literature Review
• Problem Statement
• System Requirements
• Modules to be Developed
• References
3

ABSTRACT

The proliferation of deep fake technology has raised significant


concerns regarding the authenticity of video content. This
project aims to develop a robust deep learning model capable
of detecting deep fake videos, thus contributing to the integrity
of information in media and digital communications.
4

INTRODUCTION
Deepfake Detection is the task of detecting fake videos or images that
have been generated using deep learning techniques. Deepfakes are
created by using deep learning algorithms to manipulate or replace
parts of an original video or image, such as the face of a person.
The goal of deepfake detection is to identify such manipulations and
distinguish them from real videos or images. The term” deepfake” is a
combination of” deep learning” and” fake,” pertaining to the fact that
the technology uses deep learning algorithms to produce fake media
LITERATURE REVIEW 5

Author’s & Year Title Algorithms/Methods Used Key Findings

MD Shohel Rana, Deepfake Detection • Machine learning based methods The experiment
2022 • Deep learning based methods reveals the
• Statistical measurements based DeepfakeStack
methods achieves 99.65%
• Blockchain based methods accuracy and 1.0 of
AUROC.
Arash Heidari, 2023 Deepfake Detection • Deep learning methods CNN is the most
using deep learning • CNN frequently employed
methods method in articles
(61%) and is used
in almost every
category, particularly
in image and video
deepfake detection.
Rimsha Rafique, Deepfake Detection • Error level analysis The classification is
Rahma Gantassi, And Classification • Convolutional neural network (CNN) then performed via
2023 Using Error-level • K-nearest neighbors and support vector SVM and KNN. The
Analysis And Deep machine proposed method
Learning achieved highest
accuracy of 89.5%
via ResNet18 and
6

PROBLEM STATEMENT

• As deepfake technology advances, it becomes harder to


differentiate between real and fake videos. Deepfakes use
sophisticated AI techniques, making it difficult for traditional
detection methods to keep up.
• This challenge creates a critical need for improved detection
systems that can quickly and accurately identify manipulated
content, even as the technology behind deepfakes improves.
Without effective detection, issues like misinformation, fraud,
and privacy violations can worsen.
7
SYSTEM REQUIREMENTS

Hardware:
• High-performance GPUs for training deep learning models
• Sufficient storage for large datasets (e.g., Celeb-DF, DFDC).

Software:
• Frameworks like TensorFlow, PyTorch for deep learning.
• OpenCV for video processing and image analysis.
• Dataset: High-quality datasets with labeled deepfake videos such as
the DeepFake Detection Challenge Dataset (DFDC), Celeb-DF,
FaceForensics++.
MODULES TO BE DEVELOPED 8

1. Feature Extraction Models:


Facial Landmark Detection:
HOGWILD: A distributed optimization algorithm for large-scale
machine learning.
OpenFace: A deep neural network-based facial landmark detector.
MediaPipe: A framework for building multimodal applications,
including face detection and landmark tracking.
Optical Flow:
Farneback: A dense optical flow algorithm.
TVL1: A variational optical flow method.
RAFT: A recent deep learning-based optical flow model.
Texture Analysis:
CNNs: Convolutional neural networks for extracting spatial
features.
LBPs: Local Binary Patterns for texture description.
Gabor Filters: For frequency-based texture analysis.
MODULES TO BE DEVELOPED 9

2. Deep Learning Models for Detection:


CNNs:
ResNet: Residual neural networks for deep feature learning.
VGGNet: A deep convolutional network architecture.
InceptionNet: A deep neural network architecture with parallel
convolutions.
GANs:
DCGAN: Deep convolutional generative adversarial networks.
CycleGAN: A generative adversarial network for image-to-
image translation.
RNNs:
LSTM: Long Short-Term Memory networks for sequential data
processing.
GRU: Gated Recurrent Units for efficient sequential modeling.

3. Fusion Models:
Ensemble Learning: Combining multiple models for improved accuracy.
Multimodal Fusion: Combining features from different modalities (e.g., visual,
audio).

4. Evaluation and Testing Models:


10

REFERENCES

• MD Shohel Rana, 2022. Deepfake Detection.


• Arash Heidari, 2023. Deepfake Detection using deep learning
methods.
• Rimsha Rafique, Rahma Gantassi, 2023. Deepfake Detection And
Classification Using Error-level Analysis And Deep Learning.
THANK
YOU

You might also like