Individual Source Camera Identification with Convolutional Neural Networks

• Autor(a) principal: Bernacki, Jarosław
• Data de Publicação: 2022
• Outros Autores: Costa, Kelton A. P. [UNESP], Scherer, Rafał
• Tipo de documento: Artigo de conferência
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1007/978-981-19-8234-7_4; http://hdl.handle.net/11449/246490
• topic: Camera identification; Digital forensics; Image processing; Imaging sensor identification; Security

In this paper we consider the issue of digital camera identification which matches the area of digital forensics. This problem is well-known in the literature and many algorithms based on camera’s fingerprint have been proposed. However, one may find that there is a little number of methods providing a fast and accurate digital camera identification. This problem is especially observed in terms of today’s digital cameras, producing images of big sizes. In this paper we discuss several existing approaches based on convolutional neural networks (CNN). We try to find out whether it is possible to speed up the process of learning the networks by the images. One of the findings include replacing the ReLU with SELU activation function. We experimentally show that using SELU speeds up significantly the process of learning. We also compare the identification accuracy of all considered methods. The experiments are held on extensive image dataset, consisting of many images coming from modern cameras.