A complex processing chain is applied from the moment a raw image is acquired until the final image is obtained. This process transforms the originally Poisson-distributed noise into a complex noise model. Noise inconsistency analysis is a rich source for forgery detection, as forged regions have likely undergone a different processing pipeline or out-camera processing. We propose a multi-scale approach, which is shown to be suitable for analyzing the highly correlated noise present in JPEG-compressed images. We estimate a noise curve for each image block, in each color channel and at each scale. We then compare each noise curve to its corresponding noise curve obtained from the whole image by counting the percentage of bins of the local noise curve that are below the global one. This procedure yields crucial detection cues since many forgeries create a local noise deficit. Our method is shown to be competitive with the state of the art. It outperforms all other methods when evaluated using the MCC score, or on forged regions large enough and for colorization attacks, regardless of the evaluation metric.

Images undergo a complex processing chain from the moment light reaches the camera's sensor until the final digital image is delivered. Each of its operations leaves traces on the noise model which enable forgery detection through noise analysis. In this article, we describe the Noisesniffer method [Gardella et al., Noisesniffer: a Fully Automatic Image Forgery Detector Based on Noise Analysis, IEEE International Workshop on Biometrics and Forensics, 2021]. This method estimates for each image a background stochastic model which makes it possible to detect local noise anomalies characterized by their number of false alarms. We improve on the original formulation of the method by introducing a region-growing algorithm to detect local deviations from the background model. Results show that the proposed method outperforms the previous version as well as the state of the art.

In the article 'Exposing Fake Images with Forensic Similarity Graphs', O. Mayer and M. C. Stamm introduce a novel image forgery detection method. The proposed method is built on a graph-based representation of images, where image patches are represented as the vertices of the graph, and the edge weights are assigned in order to reflect the forensic similarity between the connected patches. In this representation, forged regions form highly connected subgraphs. Therefore, forgery detection and localization can be cast as a cluster analysis problem on the similarity graph. The authors present two graph clustering methods to detect and localize image forgeries. In this paper, we present briefly the method and offer an online executable version allowing everyone to test it on their own suspicious images.

In the article 'Forensic Similarity for Digital Images', O. Mayer and M. C. Stamm introduce the forensic similarity approach, which aims at determining whether two image patches contain the same forensic traces or not. The proposed method is based on a feed-forward neural network which consists of two modules: a feature extraction module using a pair of CNNs in a siamese configuration, and a three-layer neural network that maps the extracted features into a similarity score. In this article, we explore the use of the forensic similarity score for source camera model comparison, as one of the possible applications of such an approach suggested by Mayer and Stamm.

We propose a block-based signal-dependent noise estimation method on videos, that leverages inter-frame redundancy to separate noise from signal. Block matching is applied to find block pairs between two consecutive frames with similar signal. Then the Ponomarenko et al. method is extended to video by sorting pairs by their low-frequency energy and estimating noise in the high frequencies. Experiments on a real dataset of drone videos show its performance for different parameter settings and different noise levels. Two extensions of the proposed method using subpixel matching and for multiscale noise estimation are respectively analyzed.