A Run-Length and Discrete Cosine Transform Based Technique for Image Splicing Detection

Digital images have emerged as the most popular means for sharing information in articles, newspapers, and even courtrooms. However, the widespread use of advanced digital imaging tools has made it easier to forge images. One such technique is image splicing, where multiple source images are merged into a single destination image to conceal or alter its content. Image splicing is an effective forgery technique, as it is difficult to detect by the naked eye. Detection of image splicing is a pattern recognition problem, based on finding image features that are sensitive to splicing. In this paper, we present a new image splicing detection technique based on Run-length features and discrete cosine transform (DCT). Our proposed technique uses support vector machine (SVM) to classify authentic and spliced images. We used SVM instead of generative models to reduce energy consumption and carbon footprint in machine learning applications. Experimental results show that our technique achieves a detection accuracy of 85%, outperforming other steganalysis-based techniques.