SHREE CHANAKYA EDUCATION SOCIETY’S
INDIRA COLLEGE OF ENGINEERING AND
MANAGEMENT
Approved By AICTE New Delhi, DTE (MS) and Affiliated to Savitribai Phule Pune University
(Id-No. PU/PN/Engg/282/2007)
TE SEMINAR SYNOPSIS
ON
DEEPFAKE DETECTION
SUBMITTED
BY
PAVITRA MALIGOUDA (ROLL NO.: TEB58)
Under the Guidance of
PROF. SHRADDHA SURYAWANSHI
SEM-V
YEAR 2025-2026
DEPARTMENT OF COMPUTER ENGINEERING
INTRODUCTION:
DeepFakes are AI-driven synthetic visuals that alter or mimic human faces.
These creations are often so convincing that distinguishing them from
genuine media becomes challenging. Their rise poses a serious threat to the
authenticity of online visual content.
MOTIVATION:
With DeepFakes being used for misinformation, identity misuse, and
manipulation, the need for dependable detection solutions is urgent.
Developing tools that can automatically spot such manipulations is key to
preserving digital trust.
PROBLEM DEFINITION:
The goal is to design an AI-powered system capable of identifying
DeepFake images and videos involving human faces, ensuring effective
detection across diverse creation techniques.
AIM:
To research and implement AI-based methods for identifying DeepFakes in
images and videos, supporting the reliability of multimedia sources.
OBJECTIVE:
- Examine how DeepFake content is produced.
- Study detection models, tools, and public datasets.
- Develop a detection framework using CNN and RNN architectures.
- Test detection performance and suggest improvements.
DESCRIPTION:
The system follows a sequence of steps: data preprocessing, extracting facial
features, classification, and final decision-making. Together, these stages
assess whether a media input is authentic or artificially created.
BLOCK DIAGRAM:
KEY FEATURES (Algorithms Used):
• CNNs to extract fine details from images.
• RNNs / LSTMs to analyze temporal consistency in videos.
• PRNU analysis for forensic-level detection.
• ResNet, SVM and other classifiers for final predictions.
ADVANTAGES: -
• Provides reliable detection for a wide variety of manipulated
media.
• Can be extended to real-time video monitoring.
• Useful in digital forensics and fact-checking platforms.
• Adaptable for different formats and resolutions.
LIMITATIONS: -
• Requires large, high-quality datasets for training.
• Detection performance may drop with novel DeepFake methods.
• Training deep models consumes significant resources.
• Vulnerable to adversarial manipulations.
CONCLUSION:
As DeepFakes grow more realistic and widespread, strong detection systems
are essential. This work reviews AI-based detection strategies that inspect
both visual and temporal cues. While current approaches show promise,
continued research is needed to strengthen their adaptability against
emerging DeepFake techniques.
REFERENCES: -
1.Tolosana, R., Vera-Rodriguez, R., Fierrez, J., Morales, A., &
OrtegaGarcia, J. (2020). DeepFakes and Beyond: A Survey of Face
Manipulation and Fake Detection. Information Fusion, 64, 131–148.
2. Mirsky, Y., & Lee, W. (2021). The Creation and Detection of DeepFakes:
A Survey. ACM Computing Surveys, 54(1), 1–41.
3. Wang, S.-Y., Wang, O., Zhang, R., Owens, A., & Efros, A. A. (2020).
CNN-Generated Images Are Surprisingly Easy to Spot… for Now.
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern
Recognition (CVPR), 8695–8704.
Seminar Guide Seminar Coordinator HOD
Ms Shraddha Ms Shwetkranti Taware Dr.Soumitra Das
Suryawanshi
