A Spectroscopy-Network-Based Method for Forgery Detection of Documents

Abstract

This work introduces a novel approach for forgery detection based on the spectroscopy of documents’ matters. The proposed approach uses concepts from network science to generate a weighted network of spectrums for both the original and questioned documents together. The nodes of the network represent the spectrums and the edges 15 are the correlations among them. The detection method is based on the number of clusters obtained from the tested network using our modified version of the Louvain algorithm. The spectrums of inks and papers used in printing the documents were obtained using Laser-Induced Breakdown Spectroscopy (LIBS) technology. The results of the proposed approach can be visually interpreted, which is more comfortable and 20 familiar to investigators. The proposed approach was tested under a variety of cases such as inkjet prints, laser prints, and different kinds of printing papers. The findings showed that the proposed approach always successful in distinguishing different kinds of documents with an accuracy of 100%. Finally, the proposed approach is considered simple and does not need complex computations compared to the approaches in the 25 literature.