Prototipo sistema de detección de intrusiones militar: Diseño y validación con modelo híbrido

Abstract

CONDITION FOR PUBLICATION OF PROJECT. project develops and validates an artificial intelligence model for an Intrusion Detection System (IDS) with the goal of identifying cyber threats in real time within local military networks. The justification is based on the critical need to overcome the limitations of traditional systems in the face of sophisticated attacks, proposing an autonomous and adaptable approach for high-security infrastructures. For the development, a comparative experimental evaluation was conducted between four Deep Learning architectures, using the CICIDS2017 dataset and tools such as TensorFlow and Keras. As a result of this competition, the hybrid 1D-CNN + LSTM model was selected for its superior performance, achieving an F1 score of 0.9894 and the lowest number of false negatives, the most critical error in this context. Subsequently, a functional prototype integrated this model and validated its effectiveness by successfully detecting a live port scanning attack. In conclusion, the work demonstrates the feasibility of applying hybrid AI architectures to create precise security systems, establishing a solid foundation for modernizing cyber defense.