A customized CNN model for signature authentication-Forensic implications

Area of Science:

• Computer Science
• Artificial Intelligence
• Forensic Science

Background:

• Signature authentication is crucial for verifying identity and preventing fraud.
• Traditional methods can be time-consuming and subjective.
• Deep learning offers potential for automated and accurate signature verification.

Purpose of the Study:

• To customize a deep learning-based convolutional neural network (CNN) model for signature authentication.
• To evaluate the model's performance on a dataset of genuine and forged signatures.

Main Methods:

• A convolutional neural network (CNN) model was customized and trained on 1400 signature images (700 genuine, 700 forged).
• The dataset was divided into training (1000 samples) and testing (400 samples) sets.
• Model architecture was optimized using hyperparameter tuning.

Main Results:

• The model achieved high accuracy rates: 97.32% (training), 97.92% (validation), and 84.5% (testing).
• Other performance metrics included precision (85%), recall (84%), F1-score (84%), and specificity (90%).
• The proposed model demonstrated superior performance compared to existing methods.

Conclusions:

• The customized CNN architecture provides an effective solution for signature authentication.
• The model can be further trained on larger datasets for enhanced performance.
• Potential applications include forensic document examination, banking, and legal settings.