Biometric authentication systems have gained significant importance in secure identity verification, with handwritten signatures remaining a widely accepted behavioral biometric in banking, financial, and legal domains. However, the inherent variability in individual writing styles and the prevalence of skilled forgeries make signature verification a challenging task, particularly in offline scenarios where only static images are available. This paper presents an offline handwritten signature verification system that integrates Convolutional Neural Networks (CNN) for automated feature extraction and k-Nearest Neighbors (KNN) for classification. The proposed approach leverages CNN’s capability to learn discriminative spatial features such as stroke patterns, texture, and structural characteristics directly from signature images, eliminating the need for manual feature engineering. The extracted feature vectors are subsequently classified using the KNN algorithm, which determines the authenticity of a signature based on similarity measures with stored reference samples. The performance of the system is evaluated using standard metrics such as accuracy, precision, and false acceptance/rejection rates. Experimental results demonstrate that the hybrid CNN-KNN model achieves improved classification accuracy and robustness compared to traditional methods, making it suitable for practical deployment in fraud detection systems. The study also highlights the effectiveness of combining deep learning-based feature extraction with classical machine learning classifiers for enhanced signature verification performance.