Recognition of counterfeit identity documents using FTIR spectroscopy and machine learning approaches

Area of Science:

• Forensic Science
• Analytical Chemistry
• Chemometrics

Background:

• Document forgery is a significant challenge requiring advanced analytical techniques beyond visual inspection.
• Traditional methods struggle to detect sophisticated chemical alterations in identity documents.
• Developing non-destructive methods is crucial for authenticating official documents.

Purpose of the Study:

• To present a non-destructive methodology for differentiating authentic, forged, and counterfeit Brazilian identity documents.
• To apply Fourier Transform Infrared (FTIR) spectroscopy combined with machine learning for document analysis.
• To identify chemical markers indicative of forgery and document alteration.

Main Methods:

• Analysis of 135 Brazilian identity documents (2003-2023) using FTIR spectroscopy.
• Application of Principal Component Analysis (PCA) for spectral data exploration.
• Utilisation of Partial Least Squares-Discriminant Analysis (PLS-DA) for classification and prediction.

Main Results:

• PCA identified distinct absorption bands linked to calcium carbonate and carbonyl groups, suggesting chemical erasure.
• PLS-DA models achieved high accuracy: 97.6% for the 1983 layout and 100% for the 2018 layout.
• The methodology successfully differentiated authentic, forged, and counterfeit documents at a molecular level.

Conclusions:

• FTIR spectroscopy coupled with chemometrics offers an objective, fast, and non-destructive approach for forensic document examination.
• The study provides analytical evidence of chemical strategies employed in document forgery.
• This framework advances forensic intelligence by identifying forgery patterns and criminal modus operandi.