Publicado

2026-07-01

EVALUACIÓN DE LA VULNERABILIDAD DE CAPTCHAS MEDIANTE REDES NEURONALES CONVOLUCIONALES

CAPTCHA VULNERABILITY ASSESSMENT USING CONVOLUTIONAL NEURAL NETWORKS

DOI:

https://doi.org/10.15446/rev.fac.cienc.v15n2.117839

Palabras clave:

Aprendizaje profundo , CAPTCHA , Reconocimiento de imágenes , Redes neuronales convolucionales , Seguridad informática (es)
CAPTCHA , Computer Security , Convolutional Neural Networks , Deep learning , Image recognition (en)

Descargas

Autores/as

  • Jhon Carlos Gonzales-Quilca Universidad Nacional del Altiplano
  • Elmer Ander Chalco-Huarachi Universidad Nacional del Altiplano
  • Oliver Vilca-Huayta Universidad Nacional del Altiplano https://orcid.org/0000-0002-5703-790X

La Prueba de Turing pública completamente automatizada para distinguir computadoras de humanos (en inglés CAPTCHA: Completely Automated Public Turing Test to Tell Computers and Humans Apart), es una herramienta diseñada para verificar que los usuarios son humanos y no robots, y es ampliamente utilizada para prevenir accesos no autorizados. El propósito de este artículo fue desarrollar dos modelos para el reconocimiento de CAPTCHAs de texto: El reconocimiento óptico de caracteres (OCR) y las redes neuronales convolucionales (CNN), los cuales alcanzaron una precisión del 37% y 99% respectivamente. Para entrenar y evaluar los modelos se utilizaron 10 000 imágenes generadas con la librería Gregwar. Los nuevos modelos fueron efectivos para reconocer texto distorsionado o con ruido, y las CNN demostraron robustez en el reconocimiento de CAPTCHAs de texto. En consecuencia, si se utiliza la librería Gregwar se recomienda cambiarla por otras librerías que aún son difíciles de engañar, por ejemplo, los basados en imágenes.

The Completely Automated Public Turing Test to Tell Computers and Humans Apart (CAPTCHA) is a security mechanism designed to distinguish human users from automated bots and is widely employed to prevent unauthorized access. The objective of this study was to develop two models for recognizing text-based CAPTCHAs: Optical Character Recognition (OCR) and Convolutional Neural Networks (CNNs), which achieved accuracies of 37% and 99%, respectively. A dataset comprising 10,000 CAPTCHA images generated using the Gregwar library was employed for training and evaluating the proposed models. The developed models proved effective in recognizing distorted and noisy text, while CNNs, in particular, demonstrated high robustness in text-based CAPTCHA recognition. Consequently, for systems that currently rely on the Gregwar library, it is recommended to replace it with alternative CAPTCHA libraries that remain more resistant to automated attacks, such as image-based CAPTCHA schemes.

Referencias

Abdullah Hasan, W. K. (2016). A Survey of Current Research on CAPTCHA. International Journal of Computer Science & Engineering Survey, 7(3), 1-21. https://doi.org/10.5121/ijcses.2016.7301

Bosch, A., Roma, J. C., & Bagén, A. L. (2019). Deep learning: principios y fundamentos. Editorial UOC.

Bostik, O., Horak, K., Kratochvila, L., Zemcik, T., & Bilik, S. (2021). Semi-supervised deep learning approach to break common CAPTCHAs. Neural Computing and Applications 2021 33:20, 33(20), 13333-13343. https://doi.org/10.1007/S00521-021-05957-0

Bursztein, E., Jonathan, A., Moscicki, A., & Mitchell, J. C. (2014, agosto 19). The End is Nigh: Generic Solving of Text-based {CAPTCHAs}. 8th USENIX Workshop on Offensive Technologies (WOOT 14). https://www.usenix.org/conference/woot14/workshop-program/presentation/bursztein

Dankwa, S., & Yang, L. (2021). An Efficient and Accurate Depth-Wise Separable Convolutional Neural Network for Cybersecurity Vulnerability Assessment Based on CAPTCHA Breaking. Electronics 2021, Vol. 10, Page 480, 10(4), 480. https://doi.org/10.3390/ELECTRONICS10040480

Heras, J. M. (2020). Precision, recall, F1, accuracy en clasificación. IArtificial. net.

Hu, Y., Chen, L., & Cheng, J. (2018). A CAPTCHA recognition technology based on deep learning. 2018 13th IEEE Conference on Industrial Electronics and Applications (ICIEA), 617-620. https://doi.org/10.1109/ICIEA.2018.8397789

Lin, D., Lin, F., Lv, Y., Cai, F., & Cao, D. (2018). Chinese Character CAPTCHA Recognition and performance estimation via deep neural network. Neurocomputing, 288, 11-19. https://doi.org/10.1016/J.NEUCOM.2017.02.105

Ling-Zi, X., & Yi-Chun, Z. (2012). A Case Study of Text-Based CAPTCHA Attacks. 2012 International Conference on Cyber-Enabled Distributed Computing and Knowledge Discovery, 121-124. https://doi.org/10.1109/CyberC.2012.28

Long, C., Huang, M., Ye, X., Futamura, Y., & Sakurai, T. (2025). Hybrid quantum-classical-quantum convolutional neural networks. Scientific Reports 2025 15:1, 15(1), 31780-. https://doi.org/10.1038/s41598-025-13417-1

Lv, Y., Cai, F., Lin, D., & Cao, D. (2016). Chinese character CAPTCHA recognition based on convolution neural network. 2016 IEEE Congress on Evolutionary Computation, CEC 2016, 4854-4859. https://doi.org/10.1109/CEC.2016.7744412

Noury, Z., & Rezaei, M. (2020). Deep-CAPTCHA: a deep learning based CAPTCHA solver for vulnerability assessment. arXiv preprint arXiv:2006.08296.

Puig, I., López-Cerón, M., Arnau, A., Rosiñol, Ò., Cuatrecasas, M., Herreros-de-Tejada, A., Ferrández, Á., Serra-Burriel, M., Nogales, Ó., Vida, F., de Castro, L., López-Vicente, J., Vega, P., Álvarez-González, M. A., González-Santiago, J., Hernández-Conde, M., Díez-Redondo, P., Rivero-Sánchez, L., Gimeno-García, A. Z., … Ascon, N. (2019). Accuracy of the Narrow-Band Imaging International Colorectal Endoscopic Classification System in Identification of Deep Invasion in Colorectal Polyps. Gastroenterology, 156(1), 75-87. https://doi.org/10.1053/j.gastro.2018.10.004

Reza, M. H., Sibly, M. N. B., Rabbani, S. G., Sadi, S. H., Ahamed, M. F., Shafi, F. B., Sarmun, R., & Chowdhury, M. E. H. (2026). A comprehensive review of convolutional neural networks: foundations, enhancements and applications. Neural Computing and Applications 2026 38:4, 38(4), 56-. https://doi.org/10.1007/S00521-025-11827-W

Somoza Barreiro, M., & Jiménez Gutiérrez, L. M. (2016). Reconocimiento óptico de caracteres mediante imágenes en contadores de gas. http://hdl.handle.net/20.500.14352/66041

Tharwat, A. (2021). Classification assessment methods. Applied Computing and Informatics, 17(1), 168-192. https://doi.org/10.1016/j.aci.2018.08.003

Trong, N. D., Huong, T. H., & Hoang, V. T. (2023). New Cognitive Deep-Learning CAPTCHA. Sensors 2023, Vol. 23, Page 2338, 23(4), 2338. https://doi.org/10.3390/S23042338

Villares-Jimenez, J., Quinatoa-Medina, J., Rodríguez-Galán, G., Nuñez-Agurto, D., & Santillán-Tituaña, B. (2023). Vulnerability of CAPTCHA Systems Using Bots with Computer Vision Abilities. Communications in Computer and Information Science, 1755 CCIS, 329-343. https://doi.org/10.1007/978-3-031-24985-3_24

Walia, J. S., & Odugoudar, A. (2023). Vulnerability Analysis for Captchas Using Deep Learning. 3rd IEEE International Conference on ICT in Business Industry and Government, ICTBIG 2023. https://doi.org/10.1109/ICTBIG59752.2023.10456218

Witten, I. H., Frank, E., & Hall, M. A. (2011). Data mining: Practical machine learning tools and techniques (3rd ed.). Morgan Kaufmann.

Zhang, B., Xiong, Y. J., Xia, C., & Gao, Y. (2024). Transformer-based end-to-end attack on text CAPTCHAs with triplet deep attention. Computers & Security, 146, 104058. https://doi.org/10.1016/J.COSE.2024.104058

Zhao, R., Deng, X., Wang, Y., Yan, Z., Han, Z., Chen, L., Xue, Z., & Wang, Y. (2023). GeeSolver: A Generic, Efficient, and Effortless Solver with Self-Supervised Learning for Breaking Text Captchas. Proceedings - IEEE Symposium on Security and Privacy, 2023-May, 1649-1666. https://doi.org/10.1109/SP46215.2023.10179379

Dimensions

PlumX

Visitas a la página del resumen del artículo

189

Descargas

Los datos de descarga aún no están disponibles.

Cómo citar

EVALUACIÓN DE LA VULNERABILIDAD DE CAPTCHAS MEDIANTE REDES NEURONALES CONVOLUCIONALES. (2026). Revista De La Facultad De Ciencias, 15(2), 188-202. https://doi.org/10.15446/rev.fac.cienc.v15n2.117839