| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 6.46 MB | Adobe PDF |
Autores
Resumo(s)
A deteção eficaz de intrusões em redes informáticas constitui um dos desafios mais persistente e crítico da sociedade digital contemporânea. Dada a complexidade das redes modernas, torna-se um desafio constante garantir que o tráfego e os dados armazenados fiquem protegidos contra a intercetação ilícita ou o acesso indevido por parte de terceiros. Apesar dos avanços na criptografia, a proteção da privacidade e da integridade dos dados continua a depender da capacidade de monitorização dos prestadores de serviços e das organizações.
Atualmente, as soluções de segurança mais difundidas baseiam-se em assinaturas de ataques conhecidos ou regras estáticas, exigindo uma supervisão constante por parte de administradores de sistemas. Estas técnicas apresentam a desvantagem inerente de gerarem um elevado volume de falsos positivos e de serem ineficazes perante ameaças novas ou variantes comportamentais, o que limita severamente a sua eficiência operacional. Surge, por isso, a necessidade premente de aprofundar o estudo em sistemas autodidatas que possam realizar uma análise inteligente dos dados, distinguindo com maior precisão o tráfego legítimo de ocorrências anormais.
O presente trabalho propõe o desenvolvimento e a avaliação de um motor de deteção inteligente, utilizando como base de estudo registos reais de tráfego de uma instituição de ensino superior. O foco incide na aplicação de algoritmos avançados de Machine Learning e Deep Learning de modo a mitigar o impacto de falsos alertas que sobrecarregam os analistas de segurança. Através desta abordagem, pretende-se não apenas identificar anomalias, mas também desenhar uma arquitetura conceptual que permita a integração destas deteções em ecossistemas de monitorização, promovendo uma defesa mais resiliente e autónoma.
An efficient intrusion detection in computer networks remains one of the most persistent and critical challenges of today’s digital society. Given the complexity of modern networks, it is a constant challenge to ensure that traffic and stored data are protected against illegal interception or unauthorized access. Despite advances in encryption, protecting privacy and data integrity still depends on the monitoring capacity of service providers and organizations. Currently, the most common security solutions are based on signatures of known attacks or static rules, requiring constant supervision by system administrators. These techniques have the inherent disadvantage of generating a high volume of false positives and being ineffective against new threats or behavioral variations, which severely limits their operational efficiency. Therefore, there is an urgent need to study self-learning systems that can perform an intelligent analysis of data, accurately separating legitimate traffic from abnormal events. This research aims to develop and evaluate an intelligent detection engine, based on real traffic logs from a higher education institution. The focus is on applying advanced Machine Learning and Deep Learning algorithms to reduce the impact of false alerts that overwhelm security analysts. This approach aims not only to identify anomalies but also to design a conceptual architecture to integrate these detections into monitoring ecosystems, promoting a more resilient and autonomous defence.
An efficient intrusion detection in computer networks remains one of the most persistent and critical challenges of today’s digital society. Given the complexity of modern networks, it is a constant challenge to ensure that traffic and stored data are protected against illegal interception or unauthorized access. Despite advances in encryption, protecting privacy and data integrity still depends on the monitoring capacity of service providers and organizations. Currently, the most common security solutions are based on signatures of known attacks or static rules, requiring constant supervision by system administrators. These techniques have the inherent disadvantage of generating a high volume of false positives and being ineffective against new threats or behavioral variations, which severely limits their operational efficiency. Therefore, there is an urgent need to study self-learning systems that can perform an intelligent analysis of data, accurately separating legitimate traffic from abnormal events. This research aims to develop and evaluate an intelligent detection engine, based on real traffic logs from a higher education institution. The focus is on applying advanced Machine Learning and Deep Learning algorithms to reduce the impact of false alerts that overwhelm security analysts. This approach aims not only to identify anomalies but also to design a conceptual architecture to integrate these detections into monitoring ecosystems, promoting a more resilient and autonomous defence.
Descrição
Palavras-chave
Deteção de Intrusões Machine Learning Deep Learning Segurança de Redes Anomalias. Intrusion Detection Machine Learning Deep Learning Network Security Anomalies.
Contexto Educativo
Citação
Editora
Licença CC
Sem licença CC
