Dynamic forgery signature detection using CNN and PCA

Abstract

Signature verification plays a crucial role in identity authentication, with dynamic forgeries posing a significant threat to the reliability of security systems. This work presents a robust method for identifying forged signatures by leveraging the capabilities of Convolutional Neural Networks (CNNs) combined with Principal Component Analysis (PCA). CNNs are utilized to extract deep spatial features from signature images, effectively capturing intricate patterns and anomalies that often occur in forged samples.

To enhance computational efficiency and reduce feature dimensionality, PCA is applied to the extracted features. This technique retains essential information while streamlining the processing workload. The model is trained on a dataset comprising both authentic and forged signatures, enabling it to distinguish genuine signatures from those crafted by skilled forgers with improved precision.

Experimental analysis demonstrates that the integration of CNNs for feature extraction and PCA for optimization yields high accuracy while also reducing computational demands. This approach contributes to the development of reliable and scalable signature verification systems suitable for real-world applications.