User Guide for NIST Media Forensic Challenge (MFC) Datasets
Haiying Guan, Andrew Delgado, Yooyoung Lee, Amy Yates, Daniel Zhou, Timothée N. Kheyrkhah, Jonathan G. Fiscus
NIST released a set of Media Forensic Challenge (MFC) datasets developed in DARPA MediFor (Media Forensics) project to the public in the past 5 years. More than 300 individuals, 150 organizations, from 26 countries and regions worldwide use our datasets for their research. During this period, we received many questions from users about the datasets. The questions vary from what the datasets contain, what is the definition of different kinds of metadata, how to use them, to how to modify/extract the data for their own research purposes, etc. This document aims to provide a systematical user guide for MFC datasets to answer those questions. In this document, first, an introduction about the MFC is given. Second, to provide users a deep understanding of what data and metadata are collected and how they are used in MFC datasets, a simple example with the manipulation journal graph is shown. The corresponding MFC dataset reference files of the given example are explained in detail. Third, the historical evolution of MFC datasets is introduced, and the releasable MFC datasets are summarized. Two commonly used approaches to illustrate the manipulation operation histogram of a dataset are presented. Moreover, a flexible approach for the special subset selection, selective scoring, is introduced to sample the test probes from the full test set for special task evaluation. Finally, suggestions to use MFC datasets beyond of MFC evaluations are presented for user interests.
, Delgado, A.
, Lee, Y.
, Yates, A.
, Zhou, D.
, Kheyrkhah, T.
and Fiscus, J.
User Guide for NIST Media Forensic Challenge (MFC) Datasets, NIST Interagency/Internal Report (NISTIR), National Institute of Standards and Technology, Gaithersburg, MD, [online], https://doi.org/10.6028/NIST.IR.8377, https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=931782
(Accessed December 8, 2023)