Recent advancements in image editing tools made much impact on analysis of security issues with respect to digital media domain. In specific, the forged images been uploaded to create a panic situation for the users. Those synthesized images with fake content can be used in social media, which may cause several problems. Hence, it is important for the image forensics to detect the forged or manipulated images. With the motive to effectively detect the fake images and to provide security, the proposed work focused on developing an Advanced Fake Image-Feature Network (AFIFN) based on deep learning methods. In this model, Discrete Cosine Transformation (DCT) and Y Cr Cb based image pre-processing is employed. Further, the AFIFN is framed with two-layered network structure, obtaining the pair-wise data as input. The network is trained for differentiating the features between the forged and real images. Additionally, a classification layer is added with the framed AFIFN to detect the input image is forged or not. The results show that the proposed model significantly outperforms the results of other existing models in image forgery detection.

中文翻译：

---

基于深度学习的高级假图像特征网络（AFIFN）安全模型用于图像伪造检测


图像编辑工具的最新进展对数字媒体领域的安全问题分析产生了很大影响。具体来说，伪造的图像被上传，给用户造成恐慌。这些带有虚假内容的合成图像可以在社交媒体中使用，这可能会导致一些问题。因此，图像取证检测伪造或操纵的图像非常重要。出于有效检测假图像并提供安全性的动机，拟议的工作重点是开发基于深度学习方法的高级假图像特征网络（AFIFN）。在此模型中，采用了基于离散余弦变换 (DCT) 和 Y Cr Cb 的图像预处理。此外，AFIFN采用两层网络结构，获取成对数据作为输入。该网络经过训练，可以区分伪造图像和真实图像之间的特征。此外，还添加了带有框架 AFIFN 的分类层来检测输入图像是否伪造。结果表明，所提出的模型在图像伪造检测方面显着优于其他现有模型的结果。