A Comparative Analysis of Forensic Similarity and Scale Invariant Feature Transform (SIFT) for Forensic Image Identification

The image manipulation process has contributed to the widespread dissemination of false information. image forensics can help law enforcement agencies in addressing the spread of false news or information issues through visual media. Forensic image identification can be conducted using various methods, including Scale Invariant Feature Transform (SIFT) and Forensic Similarity. This study compared two methods, SIFT and Forensic Similarity, for forensic image identification. The test results showed the SIFT method identified image forensics by detecting image similarity through calculation of the key point values of each image. The process of searching the key point values was performed to extract information from the image. A high key point value indicated a large amount of information obtained from the image extraction results. On the other hand, the Forensic Similarity method also performed image forensic detection by examining whether image patches shared the same forensic traces. The advantage of the Forensic Similarity method over the SIFT method was that Forensic Similarity was more detailed because it involved many processes. Thus, Forensic Similarity was able to find similarities between two image patch objects. Additionally, the results obtained from the Forensic Similarity method were more detailed in detecting image similarity by considering the key point matching value and Cosine Similarity. Several previous studies have already implemented the SIFT and Forensic Similarity methods for image forensics, but there was no research that directly compared these two methods. This is the strength of this research. However, this study only used three data samples from three different devices for data collection. Future research can use a larger sample size to observe the comparison results