Printer source identification by feature modeling in the total variable printer space
Roozbeh Hamzehyan1, Farbod Razzazi1, Alireza Behrad2
1Department of Electrical and Computer Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran.
Journal of Forensic Sciences
|August 25, 2021
Summary
This study introduces a novel machine learning approach for printer source identification using Local Binary Pattern (LBP) features. The method achieves high accuracy, advancing digital forensics capabilities.
Area of Science:
- Digital Forensics
- Machine Learning
- Image Analysis
Background:
- Digital forensics is rapidly evolving with advancements in digital technology.
- Printer source identification is a critical area within digital forensics.
- Existing methods may be limited by language or computational cost.
Purpose of the Study:
- To develop a novel, efficient, and language-independent method for printer source identification.
- To leverage Local Binary Pattern (LBP) features for enhanced forensic analysis.
- To reduce computational complexity in forensic document examination.
Main Methods:
- Modeling primary Local Binary Pattern (LBP) features in printer space.
- Extracting secondary features using joint factor analysis.
- Employing low-dimensional i-vector features per document image, avoiding OCR.
Main Results:
- The proposed algorithm achieved an accuracy of 98.48% in printer source identification.
- The method effectively extracts discriminant information from sparse print texture.
- The approach demonstrates comparable performance to state-of-the-art methods.
Conclusions:
- The developed method offers a robust and efficient solution for printer source identification.
- The language and character set independence broadens its applicability in digital forensics.
- This technique significantly reduces computational cost and complexity.
