Detección de patrones de Phishing en imágenes mediante técnicas de Deep Learning
Fecha
2024-12-18
Autores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Nacional Arturo Jauretche. Instituto de Ingeniería y Agronomía
Resumen
Este trabajo presenta una solución basada en técnicas de Deep Learning para detectar patrones visuales en imágenes asociadas a phishing, una de las principales amenazas en ciberseguridad. Utilizando la arquitectura VGG16 basada en redes neuronales convolucionales (CNN), se implementó y evaluó un modelo capaz de identificar sitios maliciosos, alcanzando una precisión cercana al 90%.
El estudio destaca la importancia de contar con conjuntos de datos balanceados para evitar sesgos y mejorar la capacidad de generalización del modelo. Durante la investigación se abordaron desafíos como la detección de imágenes ambiguas y las limitaciones computacionales para realizar el entrenamiento.
El modelo desarrollado representa un aporte en la lucha contra el phishing y sienta las bases para futuras mejoras, como la ampliación del dataset, la optimización del modelo y su implementación en sistemas de tiempo real.
This work presents a Deep Learning-based solution for detecting visual patterns in images associated with phishing, one of the most prevalent threats in cybersecurity. Using the VGG16 convolutional neural network (CNN) architecture, a model was implemented and trained to identify malicious sites by analyzing their visual characteristics. The results achieved demonstrate a precision of nearly 90%, validating the effectiveness of the proposed approach. The study highlights the importance of using balanced datasets to improve model generalization and reduce prediction bias. While the model performed well on complex phishing images, challenges were identified, particularly in handling ambiguous images visually similar to legitimate websites and in computational resource limitations. In conclusion, the developed model represents an advancement in phishing detection. Future work includes expanding the dataset, optimizing the model with advanced architectures, and implementing real-time detection systems to address evolving cyber threats.
This work presents a Deep Learning-based solution for detecting visual patterns in images associated with phishing, one of the most prevalent threats in cybersecurity. Using the VGG16 convolutional neural network (CNN) architecture, a model was implemented and trained to identify malicious sites by analyzing their visual characteristics. The results achieved demonstrate a precision of nearly 90%, validating the effectiveness of the proposed approach. The study highlights the importance of using balanced datasets to improve model generalization and reduce prediction bias. While the model performed well on complex phishing images, challenges were identified, particularly in handling ambiguous images visually similar to legitimate websites and in computational resource limitations. In conclusion, the developed model represents an advancement in phishing detection. Future work includes expanding the dataset, optimizing the model with advanced architectures, and implementing real-time detection systems to address evolving cyber threats.
Descripción
Palabras clave
Aprendizaje profundo, Phishing, Redes Neuronales Convolucionales, VGG16, Ciberseguridad, Detección de patrones visuales, Deep Learning, Convolutional Neural Networks, Cybersecurity, Visual Pattern Detection
Citación
Soria, K. M. (2024). Detección de patrones de Phishing en imágenes mediante técnicas de Deep Learning [Práctica Profesional Supervisada, Universidad Nacional Arturo Jauretche]. https://rid.unaj.edu.ar/handle/123456789/3305