Analysis of Periodic Marks for Forensic Document Examination using Boosting Algorithms

Abstract

The increase in illicit activities, particularly financial crimes like bank fraud, tax evasion, counterfeit goods, and fraudulent schemes, is having an impact on the national economy. In instances of such occurrences, it is imperative for a scientist to meticulously assess the pertinent documents in order to retain their evidential significance through non-destructive techniques. The 2D Lab Formation Sensor allowed for the differentiation of various types of copy paper through a non-destructive method. In this study, machine learning technologies were combined with a 2D Lab Formation Sensor to identify document paper for forensic purposes. Adaptive Boost, Gradient Boosting Machine, and Extreme Gradient Boost models were built. The XGBoost model outperformed the others, achieving an accuracy of 0.88. The periodic marks from the forming fabric serve as unique characteristics of each copy paper, depending on the manufacturer.