Evolución gramatical antagónica para la ciberseguridad
Loading...
Official URL
Full text at PDC
Publication date
2024
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
En ciberseguridad, es vital mejorar la defensa de redes ante crecientes amenazas, lo cual requiere comprender el impacto de tácticas sofisticadas en dispositivos y sistemas interconectados. Microsoft desarrollo CyberBattleSim, una herramienta que simula escenarios empresariales reales para analizar y crear estrategias de defensa, aunque necesita una configuración compleja. El proyecto de fin de grado propone integrar mecanismos para realizar experimentos de forma automatizada y presenta un primer experimento que cambia la forma de generar la estructura de la red de nodos en CyberBattleSim. Este experimento evalúa cómo un modelo antagónico basado en evolución gramatical, que genera redes de nodos mediante reglas de ensamblaje y programación evolutiva, afecta la eficacia del atacante. El modelo producido por coevolución antagónica selecciona redes ´óptimas para dificultar las acciones maliciosas, destacando la importancia de la topología de la red en seguridad y proponiendo una defensa basada en acciones pasivas. La automatización del entrenamiento sobre conjuntos de datos permite evaluar y analizar resultados más eficientemente, eliminando la necesidad de interacción manual
In the realm of cybersecurity, it is essential to enhance network defenses against increasing threats, which requires understanding the impact of sophisticated tactics on interconnected devices and systems. Microsoft developed CyberBattleSim, a tool that simulates real enterprise scenarios to analyze and develop defense strategies, although it necessitates complex configuration. The degree final project proposes integrating mechanisms for automated experiments and presents an initial experiment that alters the node network structure in CyberBattleSim. This experiment assesses how an adversarial model based on grammatical evolution, which generates node networks through assembly rules and evolutionary programming, affects the attacker’s effectiveness. The coevolutionary approach selects optimal networks to hinder malicious actions, highlighting the importance of network topology in security and proposing a defense based on passive actions. Automating the training on datasets allows for more efficient evaluation and analysis of results, eliminating the need for manual interaction.
In the realm of cybersecurity, it is essential to enhance network defenses against increasing threats, which requires understanding the impact of sophisticated tactics on interconnected devices and systems. Microsoft developed CyberBattleSim, a tool that simulates real enterprise scenarios to analyze and develop defense strategies, although it necessitates complex configuration. The degree final project proposes integrating mechanisms for automated experiments and presents an initial experiment that alters the node network structure in CyberBattleSim. This experiment assesses how an adversarial model based on grammatical evolution, which generates node networks through assembly rules and evolutionary programming, affects the attacker’s effectiveness. The coevolutionary approach selects optimal networks to hinder malicious actions, highlighting the importance of network topology in security and proposing a defense based on passive actions. Automating the training on datasets allows for more efficient evaluation and analysis of results, eliminating the need for manual interaction.













