Z-ENG: DeepFake detection

2024-2025 tavasz

Szoftver

Téma leírása

Introduction Deepfake technology, powered by artificial intelligence, has seen significant advances in recent years. By leveraging deep learning models such as Generative Adversarial Networks (GANs), deepfakes allow for the creation of highly realistic fake videos, images, and audio. While these innovations have positive applications in fields like entertainment and education, they also present serious ethical and security challenges. Malicious uses of deepfakes include misinformation campaigns, identity theft, and reputational damage. The ability to detect deepfakes is therefore critical in preserving trust and authenticity in digital media.

Main Areas and Techniques in Deepfake Detection: 

  1. Image and Video Analysis

  • Identifying inconsistencies in lighting, shadows, or reflections.
  • Detecting irregularities in facial movements or blinking patterns.
  • Analyzing frame-by-frame discrepancies in video content

   2. Audio Analysis

  • Detecting manipulated voice patterns.
  • Identifying inconsistencies in pitch, tone, and cadence.
  • Using spectral analysis to distinguish real from synthesized audio

  3. Machine Learning Approaches

  • Training convolutional neural networks (CNNs) for image-based detection.
  • Leveraging recurrent neural networks (RNNs) for sequential video or audio data.
  • Using ensemble learning to combine multiple detection methods.

  4. Watermark Detection

  • Exploring methods for embedding and detecting invisible watermarks in media.
  • Developing techniques to identify tampered or removed watermarks.
  • Using watermarks to trace the origin of manipulated content.


Students may choose to focus on one or more sections of this project based on their interests and skills. The work can take one of two primary directions:

  1. Research and Algorithm Evaluation

  • Implement and evaluate different algorithms for deepfake detection.
  • Perform comparative studies to analyze their strengths, weaknesses, and performance metrics.

   2. Software Development

  • Build a practical application (web or mobile) that implements deepfake detection techniques.
  • Design user-friendly interfaces to allow users to upload media and receive real-time feedback on its authenticity.
  • Ensure scalability and robustness for deployment in real-world scenarios.

The department is working on a relevant project, you can take advantage of existing work (researched algorithms, tools, ..etc) if necessary.

If you are interested in this topic and believe you have the necessary skills, please get in touch via email or Teams. Include the following information: your name, whether you are at the BSc or MSc level, your relevant skills, and the specific topic you wish to work on with a brief description of your proposed approach.

Feltételek

  • machine learning

Maximális létszám: 8 fő