Call for Papers: Deep Learning Algorithms and Techniques to Identify Deepfakes

Recent years have witnessed the rise and widespread use of deep learning techniques in a variety of areas, ranging from simple data analysis to complex image classification tasks. Lately, deep learning techniques have also been employed to produce “deepfakes” – digital content that synthesizes images with other audio and videos to create altered footage.

The widespread use of image sharing social media platforms provides large amounts of data that can be combined with deep learning techniques, especially generative adversarial networks (GANs), to produce deepfakes that appear authentic. An alarming side of deepfakes is the potential to create misleading and/or explicit content, with predominantly influential personalities like celebrities, politicians, and even religious leaders being targeted. Deepfakes not only erode and ruin people’s trust, but they can also have financial implications – for example, a deepfake about a CEO or director could destroy an entire company. Future deepfakes could feasibly be used to create political distress, blackmail, or even fake terrorism events that did not happen.

Deep learning-based networks, including GANs, convolutional neural networks (CNNs), recurrent neural networks (RNNs) and deep neural networks (DNNs) are often employed to generate deepfakes, with conventional image forgery detection techniques time and again unable to identify them. Blockchain and distributed ledger technologies (DLTs) have shown promising results, and could be employed to combat deepfakes in the future. However, further research is needed to develop techniques for the identification of most dangerous types of deepfakes, to enhance people’s trust in media content and avoid social and financial disturbances.

Keywords: Deep learning, Authentication, Video evidence, Blockchain, Forgery detection, Deep learning algorithms and architectures, Reinforcement learning.

The scope of the collection includes, but is not limited to, the following topics:

  • Deep Learning techniques for deepfakes detection
  • Blockchain approaches to identify deepfakes
  • Distributive ledger techniques
  • Full face synthesis techniques
  • Identity swap techniques
  • Deep learning techniques based on DNNs, CNNs, RNNs, LSTMS, GANs, etc.
  • Advance hybrid approaches for deepfakes
  • Face manipulation techniques
  • Media forensic techniques
  • Ethereum based techniques for deepfakes
  • Dapps techniques and algorithms
  • Deep learning architectures
  • Forgery detection techniques
  • Pairwise learning to identify deepfakes
  • Techniques to verify PoA

Collection Editors

Dr. Imran Ashraf. Department of Information & Communication Engineering, Yeungnam University, South Korea.

Dr. Ali Kashif Bashir. Department of Computing and Mathematics, Manchester Metropolitan University, United Kingdom.

Dr. Yousaf Bin Zikria. Department of Information & Communication Engineering, Yeungnam University, South Korea.

Editorial Oversight

Dr. Arkaitz Zubiaga. Social Data Science lab, Queen Mary University of London, United Kingdom.

 

Submission Guidelines

PeerJ Computer Science is currently accepting submissions for this Special Issue, with a submission deadline of 2nd September, 2021. Submit Now.

When submitting please add a note stating “Submitted to the Identifying Deepfakes Special Issue” in the Confidential Note to Staff field of the submission form. Authors should review PeerJ’s submission instructions before submitting. All submitted papers will be peer reviewed using the normal standards of PeerJ Computer Science.

Unpublished manuscripts or extended versions of the papers that have been published in related conference proceedings are welcome. All submissions must be original and not currently under review for publication elsewhere. Conference papers may only be submitted if the paper was completely re-written or substantially extended (50%).

Authors should note that, if accepted for publication, they will be required to pay either an Article Processing Charge ($1195) or for a PeerJ Lifetime Membership (a one-off payment for lifetime publishing privileges – starting at just $399).

 

You may also like...