International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024
p-ISSN: 2395-0072
www.irjet.net
A SURVEY ON DEEPFAKE VIDEO DETECTION USING DEEPLEARNING Mrs.M.Lakshmi Prabha1, N.Aakash2, S.Balaganesh3, B.Roopan4 1Assosiate Professor, Dept. of IT, Sri Manakula Vinayagar Engineering College, Madagadipet, Puducherry 2,3,4 Student, Dept. of IT, Sri Manakula Vinayagar Engineering College, Madagadipet, Puducherry
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - In response to the growing danger that deepfake
threat of deepfake videos and restore trust and authenticity in digital media.
films represent to the authenticity of visual material, this study explores the effectiveness of using physiological signals heart rate, eye movement, and facial emotions in particular for detection. The identification of falsified content is significantly hampered by deepfake technology, which has led to research into new detection techniques. Using knowledge from other studies, we examine whether physiological signals can serve as trustworthy markers of dishonesty. Promising approaches for identifying irregularities in visual content include heart rate monitoring, eye movement tracking, and facial expression detection. By doing an extensive analysis of current research, techniques, and datasets, we evaluate the benefits and drawbacks of each strategy, opening the door for hybrid models that combine several physiological signals for improved accuracy. ( Size 10 & Italic , cambria font)
1.1 HEART RATE ANALYSIS Heart rate serves as a fundamental physiological signal, representing the frequency of heartbeats per minute. In video analysis, heart rate can be measured through various methods, including photoplethysmography (PPG) sensors or computer vision techniques that track changes in facial blood flow. Heart rate variability (HRV), the variation in time intervals between consecutive heartbeats, holds particular significance in detecting deception. Research suggests that heightened emotional arousal, often associated with deception, can influence HRV patterns, making it a potential marker for identifying deceitful behavior.
1.2 EYE MOVEMENT ANALYSIS
Key Words: CNN, RNN, LSTM, PPG, GNN
Eye movement patterns provide valuable insights into cognitive processes and attentional mechanisms. Eyetracking technology enables precise measurement and analysis of eye movements, including fixations, saccades, and smooth pursuit. In video analysis, eye-tracking data can reveal where and how individuals direct their visual attention. Correlations between eye movements and cognitive processes, such as memory retrieval and decisionmaking, underscore the utility of eye movement analysis in understanding human behavior and detecting deception.
1.INTRODUCTION The rise of deepfake technology has sparked widespread concern due to its ability to create highly convincing fake videos, blurring the line between reality and fiction. These videos, generated using sophisticated AI algorithms, can manipulate images and audio to make individuals appear to say or do things they never did. This poses a serious threat to the credibility of visual media and raises questions about the reliability of information shared online. Traditional methods of detecting deepfakes, such as manual inspection, are often inadequate in the face of rapidly advancing technology. In response, researchers are exploring new avenues, including the analysis of physiological signals, to detect these deceptive videos. Physiological signals like heart rate, eye movement, and facial expressions can provide valuable insights into how our bodies react to what we see. By studying these signals, researchers hope to uncover subtle differences between real and fake videos, ultimately developing automated tools for deepfake detection. This survey paper aims to explore the use of physiological signals, along with advanced deep learning techniques like Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks, in the fight against deepfakes. By examining existing research, methodologies, and challenges, we seek to provide a comprehensive understanding of the current state of deepfake detection. Additionally, we aim to identify areas for improvement and propose future research directions to enhance the effectiveness of deepfake detection methods. Ultimately, our goal is to address the growing
© 2024, IRJET
|
Impact Factor value: 8.226
1.3 FACE EXPRESSION ANALYSIS Facial expressions play a crucial role in human communication, conveying a wide range of emotions and intentions. In video analysis, facial expression recognition techniques utilize computer vision algorithms to detect and classify facial expressions based on key facial landmarks and muscle movements. Emotional cues, such as changes in facial expressions, can provide valuable cues for detecting deception. Additionally, micro-expressions brief, involuntary facial expressions that occur within milliseconds hold particular promise in discerning concealed emotions and deceptive behavior in video content.
1.4 CNN BASED APPROACHES CNN-based approaches are making significant strides in the realm of deepfake detection. Researchers are harnessing powerful architectures like VGGNet, ResNet, and InceptionNet to sift through visual data and pinpoint telltale
|
ISO 9001:2008 Certified Journal
|
Page 1340