This study proposes a hybrid CNN-LSTM model optimized by the Artificial Bee Colony (ABC) algorithm for handwritten signature recognition. Addressing the limitations of traditional methods in capturing dynamic signature variations, the framework combines CNN’s spatial feature extraction with LSTM’s sequential pattern analysis. Hyperparameters (LSTM units, dropout rate, learning rate) are automatically tuned via ABC, achieving 100% accuracy and an F1-score of 1.0 on a dataset of 2,000 signatures (10 participants, 100 genuine and forged samples each). Preprocessing steps including resizing (128 × 128 pixels), normalization, and augmentation enhance generalizability. The confusion matrix and training curves validate the model’s low error rates and stability. This work demonstrates the efficacy of bio-inspired optimization in deep learning for biometric security, offering a scalable solution for real-world applications.