A Universal Attack Against Histogram-Based Image Forensics

Abstract

In this paper the authors propose a universal image counter-forensic scheme that contrasts any detector based on the analysis of the image histogram. Being universal, the scheme does not require knowledge of the detection algorithms available to the forensic analyst, and can be used to conceal traces left in the histogram of the image by any processing tool. Instead of adapting the histogram of the image to fit some statistical model, the proposed scheme makes it practically identical to the histogram of an untouched image, by solving an optimization problem. In doing this, the perceptual similarity between the processed and counter-attacked image is preserved to a large extent. The validity of the scheme in countering both contrast-enhancement and splicing- detection is assessed through experimental validation.