Network Science as a Forgery Detection Tool in Digital Forensics

Abstract

Forgery detection of documents is considered a challenging task in the field of digital forensics. The detection process is usually complex and needs a lot of stages, which consumes time and effort. The limitation of the literature lies in providing methods that can be efficiently and easily adopted with minimum cost. This work proposes a novel approach for detecting counterfeit/ forged documents. The proposed approach is based on network science approaches for analyzing documents’ ink spectrums aiming to detect whether a document was counterfeited or forged. To this end, Laser-Induced Breakdown Spectroscopy (LIBS) is used to retrieve the spectrums of the original and questioned documents. The extracted spectrums are formalized to create a dataset, which contains nodes (spectrums) and edges (correlations among spectrums). Then, the dataset is used to generate a network of spectrums that represented both the original and questioned documents. After that, the generated network is visualized and clustered. The detection process is mainly based on the information provided by network clusters (e.g., number of clusters). The results showed that the proposed approach was efficient in distinguishing documents. Moreover, the proposed approach was able to distinguish a document itself whether it was counterfeited by extracting the clusters of the questioned document.