Scholarlink

<p>The recent development and growing adoption of very superior deepfake technology have resulted in an unparalleled challenge for communication security systems in the modern digital era for security and authenticity validation purposes. A novel hybrid deep learning model architecture incorporating the benefits of Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) for real-time deepfake detection in video communication systems will be presented in this article for future development purposes. The proposed deepfake detection system in this article efficiently leverages and utilizes the superior attributes of both CNNs for the local feature extraction capability and ViTs for understanding the spatial relationships for an incredible accuracy that reaches 98.33%, precision ratio measurements of 98.97%, and 99.70% for the MIT DeepFake Detection (DFD) Dataset while meeting the total end-to-end system delay constraints of 111–127&#xa0;ms for face detection and transmission delay constraints for real-time systems. Its usability in Web Real-Time Communication (WebRTC)-based video communication systems asserts and verifies its practicability and feasibility for serving as an optimal and most effective technological solution for peer-to-peer secure digital communication purposes through a Flask-based Application Programming Interface (API).</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid deep learning architecture for real-time deepfake detection in WebRTC-based video platforms

  • 期刊论文

  • 发表日期：

    2026年3月24日

查看全文
  • Nishant,
  • Prakhar Mishra

摘要

The recent development and growing adoption of very superior deepfake technology have resulted in an unparalleled challenge for communication security systems in the modern digital era for security and authenticity validation purposes. A novel hybrid deep learning model architecture incorporating the benefits of Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) for real-time deepfake detection in video communication systems will be presented in this article for future development purposes. The proposed deepfake detection system in this article efficiently leverages and utilizes the superior attributes of both CNNs for the local feature extraction capability and ViTs for understanding the spatial relationships for an incredible accuracy that reaches 98.33%, precision ratio measurements of 98.97%, and 99.70% for the MIT DeepFake Detection (DFD) Dataset while meeting the total end-to-end system delay constraints of 111–127 ms for face detection and transmission delay constraints for real-time systems. Its usability in Web Real-Time Communication (WebRTC)-based video communication systems asserts and verifies its practicability and feasibility for serving as an optimal and most effective technological solution for peer-to-peer secure digital communication purposes through a Flask-based Application Programming Interface (API).