Detecting image splicing in the wild (WEB)

As grassroots and social media-based journalism becomes more widespread, the need to verify information coming from such channels becomes imperative. In the past, there have been multiple occasions where forged pictures successfully passed as original news items, spreading misinformation or even panic. In this work, we investigate the potential for applying today's state of the art in image splicing detection in the context of images on the Web and images disseminated through social media. We investigate the alterations social media platforms apply on images and evaluate their impact on tampering detection. We further present a real-world dataset of forged images collected from various Web sources, and attempt to evaluate them using the current state-of-the-art in splicing detection. We present our results, and discuss their implications in real-world verification settings.