Anomaly-Based Manipulation Detection in Satellite Images.

János Horváth, David Guera, Sri Kalyan Yarlagadda, Paolo Bestagini, Fengqing Maggie Zhu, Stefano Tubaro, Edward J. Delp

Abstract: Satellite overhead imagery can be easily acquired and shared. The integrity of these type of images cannot longer be assumed, due to availability of sophisticated classical and machine learning based image manipulation tools. In this paper we proposed a deep learning based method for detecting and localizing splicing manipulations in overhead images. Our method uses recent advances in anomaly detection and does not require any prior knowledge of the type of manipulations that an adversary could insert in the satellite imagery. We compare our method against robust satellite-based manipulation detection approaches. We show that our proposed technique outperforms all previous methods, especially in detecting small-sized manipulations.