Deep Learning in Image and Video Forensics

  in Special Issue   Posted on August 13, 2017

Information for the Special Issue

Special Issue Call for Papers:

The pervasiveness of new technologies, such as smartphones, tablets and Internet made digital images and videos the primary source of visual information in nowadays society. However, their reliability as a true representation of reality cannot be taken for granted, due to the affordable powerful graphics editing software that can easily alter the original content without any visual trace of the modification.

Nowadays, machine learning techniques and, in particular, Deep Learning have come to play a vital role to deal with a massive amount of unsupervised data. In recent years, deep neural networks, such as deep belief network, deep autoencoder and convolutional neural network (CNN), have shown to be capable of extracting complex statistical features and efficiently learning their representations, allowing it to generalize well across a wide variety of computer vision tasks, including image classification, speech recognition and so on.

The extensive use of Deep Learning in many areas has motivated and led the multimedia forensics community to comprehend if such technological solution is able to detect image and video manipulations or to exploit source identification. For example, it has been foreseen the proposal of new convolutional network architecture capable of working on any kind of different image formats and of automatically learning manipulation detection features directly from the training data itself. A general data-driven forensics methodology should be devised to accomplish the forensics tasks, independent from the kind of tampering, from the image format and designed to detect many, if not all, editing operations.

Furthermore, it is interesting to investigate the adversarial actions performed on deep learning techniques, to understand how their analysis can be biased and perturbed by means of the injection of fake data or adversarial examples and how the trustworthiness of the produced knowledge is diminished in relation with the kind and intensity of the performed manipulation (e.g. forged images and videos).

The aim of this special issue is to gather image forensic works specifically oriented to deal with deep learning based approach with applications in passive image forensics.

We solicit high-quality original research papers as well as review papers that mainly address these issues and advance the development in image forensics. Submitted papers should not be previously published or be under consideration for publication elsewhere.

Potential topics include, but are not limited to:

  • tampering detection machine learning classification
  • deep learning tampering detection
  • source identification with deep learning
  • relationship between adversarial forensics and deep learning

Proposed Dates

  • Submissions Oct 2nd, 2017.
  • 1st. Round of Reviews Nov 27th, 2017
  • Authors Revised Versions Jan 8th, 2018
  • 2nd Round of Reviews Feb 8th, 2018
  • Authors Revised Versions March 9th, 2018
  • Expected Decision April 2nd, 2018
  • Expected Publication May 2018

Submission Guidelines

All manuscripts and any supplementary material should be submitted through the Elsevier Editorial System (EES) at Authors should indicate that their submission is intended for this special issue on “DL-IVF” when they reach the “Article Type Name” step in the submission process. A detailed submission guideline is available at the Guide for Authors (

Lead Guest Editor

Roberto Caldelli,
CNIT – National Interuniversity Consortium for Telecommunications
MICC – Media Integration and Communication Center,
University of Florence, Florence, Italy.

Guest Editors

Marc Chaumont,
Nımes University, France.

Chang-Tsun Li
School of Computing and Mathematics,
Charles Sturt University, Australia.

Irene Amerini,
MICC – Media Integration and Communication Center,
University of Florence, Florence, Italy.

Closed Special Issues