Quantum-inspired K-nearest neighbors classifier for enhanced printer source identification in forensic document analysis

Document source identification in printer forensics focuses on determining the source printer of a document by analyzing characteristics such as printer model, serial number, defects, or unique artifacts. This is crucial in forensic investigations involving counterfeit documents or anonymous threats. However, identifying consistent patterns across different printers remains challenging, especially when perpetrators attempt to obscure these artifacts. Machine learning models in this field must identify discriminative features that differentiate printers while minimizing noise. In particular, choosing an appropriate distance metric for K-Nearest Neighbors (KNN) classifiers is critical and requires experimentation. This study proposes a quantum-inspired approach to improve KNN's performance in printer source identification. By exploring alternative number of neighbors (K), quantum-inspired computing can optimize feature space calculations, even in noisy conditions. This allows the system to iteratively refine and select the optimal K value based on classification performance, ensuring that the best K is identified for the specific dataset and task. The system utilizes the Grey Level Co-occurrence Matrix (GLCM) for feature extraction, which is robust to changes in rotation and scale. Experimental results demonstrate that the Quantum-inspired KNN (QKNN) classifier outperforms classical KNN, achieving higher accuracy in identifying subtle printing artifacts, even under variable conditions.