An Image is Worth 1000 Frauds – Detecting Fake Images and Videos presented at Kiwicon 2014

by Robert Winkel,

Summary : Magazines with photo shoots of super slim super models... Dictators releasing images that show their military might... YouTubers releasing unbelievable videos of UFOs... Presidents providing copies of their birth certificates to prove they weren’t born overseas... “Amateur” videos capturing amazing stunts using construction site tools... Hundreds of people claiming they won the $100 million lotto, with photos to prove it... Professionals displaying a copy of their credentials, certifications, or qualifications online...
With the prevalence of CGI, Photoshop, and photo and video sharing sites, we are exposed to more and more images and videos where we are doubtful of their authenticity.
This presentation will cover the more popular techniques to uncover fake images and videos. Topical, humorous, and even Kiwicon-related images and videos will be demoed. Techniques covered include:
Finding the original, unaltered image, e.g. through Google reverse image search.
Eyeballing “obvious” Photoshopped images.
Examining metadata.
Statistical methods, e.g. histogram analysis, error-level analysis, principle component analysis, wavelet transformations.
Tracing objects and their shadows to detect light source inconsistencies.
Hand-held camera jitter analysis in videos to detect unhuman-like motion.
Motion tracking of moving objects within videos to detect violations of Newton’s laws.