Behavior Knowledge Space-Based Fusion for Copy–Move Forgery Detection

Abstract— The detection of copy–move image tampering is of paramount importance nowadays, mainly due to its potential use for misleading the opinion forming process of the general public. In this paper, we go beyond traditional forgery detectors and aim at combining different properties of copy–move detection approaches by modeling the problem on a multiscale behaviour knowledge space, which encodes the output combinations of different techniques as a priori probabilities considering multiple scales of the training data. Afterward, the conditional probabilities missing entries are properly estimated through generative models applied on the existing training data. Finally, we propose different techniques that exploit the multi-directionality of the data to generate the final outcome detection map in a machine learning decision-making fashion. Experimental results on complex data sets, comparing the proposed techniques with a gamut of copy–move detection approaches and other fusion methodologies in the literature, show the effectiveness of the proposed method and its suitability for real-world applications. This first contribution is key for combining forgery detection algorithms, as oftentimes they are complementary but it is very hard to find enough training examples to cover all cross-effects of their combinations. Second, we incorporate expert knowledge to the adopted BKS representation in order to be more robust to some common operations in image tampering that can lead to confusion in the classification of individual classifiers, such as resizing and noise addition. For that, we propose a Multiscale Behavior Knowledge representation, which takes into account different scales of training data. < final year projects >