Blind Forensic Method Based on Convolutional Neural Networks for Image Splicing Detection

Abstract

In order to effectively detect whether digital images are spliced, a blind forensics method of digital image splicing based on deep learning is proposed. The method uses high-pass filter to preprocess the image, weakens the negative influence of image content on tampering forensic analysis, implements feature selection and classification based on convolutional neural networks(CNNs) to realize the classification of real images and spliced images. Experiments on the Columbia image splicing detection evaluation dataset and comparison with traditional forensic methods show that the proposed method can achieve better detection accuracy.