Image splicing detection with principal component analysis generated low-dimensional homogeneous feature set based on local binary pattern and support vector machine

Debjit Das, Ruchira Naskar, Rajat Subhra Chakraborty

Abstract: Due to the wide availability of various ready-to-use image tampering software tools, image manipulation has become widespread nowadays. Tampered images are often used intentionally for unlawful and malicious purposes. One of the most common forms of image manipulation today includes the Image Splicing attack. Image splicing can be defined as generating a composite image by combining parts of different images from different sources, forming one artificially generated composite image. Given its prevalence, image splicing detection has a fundamental importance in digital forensics, and therefore, several research works have been carried out in the recent past to address this problem. In this paper, we propose an image forensic technique using a homogeneous image feature set of optimal dimensions for the successful detection of image splicing attacks. The Columbia Image Splicing Detection Evaluation Dataset has been used in our experiments, and different feature sets extracted from the available data have been adopted and experimented with. We have succeeded in finding a feature set of dimension as low as 31 comprising of Local Binary Pattern features for image splicing detection, obtaining state-of-the-art spliced image classification result using the Support Vector Machine classifier.