Training Data Improvement for Image Forgery Detection using Comprint

TL;DR

This paper investigates how different training data compositions affect Comprint's ability to detect manipulated images, finding that recompression enhances robustness while low-quality compression does not significantly impact accuracy.

Contribution

It demonstrates that including recompressed images in training improves Comprint's robustness against image manipulations.

Findings

Recompression in training boosts detection robustness.

Low-quality compression during training has minimal impact.

Comprint can be effectively used on smartphones for image verification.

Abstract

Manipulated images are a threat to consumers worldwide, when they are used to spread disinformation. Therefore, Comprint enables forgery detection by utilizing JPEG-compression fingerprints. This paper evaluates the impact of the training set on Comprint's performance. Most interestingly, we found that including images compressed with low quality factors during training does not have a significant effect on the accuracy, whereas incorporating recompression boosts the robustness. As such, consumers can use Comprint on their smartphones to verify the authenticity of images.