Scholarlink

The accelerated development of artificial intelligence and generative models has caused the spread of deepfakes, highly realistic synthetic media that can trick observers by manipulating audio, video, or images. Although these technologies hold creative promise, they also carry tremendous risks, such as misinformation, identity theft, and loss of trust in digital content. This research paper introduces a new deepfake detection system in order to overcome these challenges using state-of-the-art machine learning approaches. Our suggested framework marries multimodal feature extraction, which unites spatial-temporal analysis of frames in the video with synchronization cues between audio and visual signals, to detect minor inconsistencies that reflect synthetic manipulation. We utilize a deep learning architecture that combines convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to learn from a diverse dataset of real and manipulated media.

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

A Deepfake Detection Using Convolutional Neural Networks (CNN): A Robust Spatial Feature-Based Approach

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Rohit Mishra,
  • Amit Kumar Tiwari,
  • Anushika Srivastava,
  • Priya Tripathi,
  • Aditya Pratap Singh

摘要

The accelerated development of artificial intelligence and generative models has caused the spread of deepfakes, highly realistic synthetic media that can trick observers by manipulating audio, video, or images. Although these technologies hold creative promise, they also carry tremendous risks, such as misinformation, identity theft, and loss of trust in digital content. This research paper introduces a new deepfake detection system in order to overcome these challenges using state-of-the-art machine learning approaches. Our suggested framework marries multimodal feature extraction, which unites spatial-temporal analysis of frames in the video with synchronization cues between audio and visual signals, to detect minor inconsistencies that reflect synthetic manipulation. We utilize a deep learning architecture that combines convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to learn from a diverse dataset of real and manipulated media.