Sistemas de detecção de intrusão com técnicas de inteligência artificial

• Ano de defesa: 2011
• Autor(a) principal: Silva, Jacson Rodrigues Correia da
• Orientador(a): Goulart, Carlos de Castro
• Banca de defesa: Correia, Luiz Henrique Andrade
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Mestrado em Ciência da Computação
• Departamento: Metodologias e técnicas da Computação; Sistemas de Computação
• País: BR
• Palavras-chave em Português: Sistemas de detecção de intrusos; Invasão; Segurança; Inteligência artificial; Ataques; SDI
• Palavras-chave em Inglês: Intrusion detection systems; Invasion; Security; Artificial intelligence; Attacks; IDS
• Área do conhecimento CNPq: CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://locus.ufv.br/handle/123456789/2621

Due the increase of the amount of important information on computer networks, security has became primordial to ensure the integrity, confidentiality and availability of data traffic. To improve security, there are useful tools such as Firewalls and Intrusion Detection Systems (IDS). Currently, methods of Artificial Intelligence (AI) are used to improve these tools. This work proposes to evaluate the improvement of the hit rates of Intrusion Detection Systems using some Artificial Intelligence techniques. They are presented the concepts of security and measures required to implement it, as well as some information about computer systems intruders. It is also presented the main concepts abow Intrusion Detection Systems and some Artificial Intelligence techniques and a review about IDS implemented with AI techniques. The development of this work has begun with the implementation of an Artificial Neural Network, which had its characteristics modified by Genetic Algorithms to improve its hit rates on normal and abnormal connections of a computer network. It was also implemented a Fuzzy System, which was then combined with the Artificial Neural Networks to create a Hybrid Intelligent System. After the implementations, the hit rates for detected intrusions and normal traffic were obtained and compared. The results showed that we have obtained higher hit rates for all methods when compared with the initial system. For each implemented method, the results were presented considering their hit rates using comparative charts with the previously implemented methods and a discussion about each new result. The correct classification of attacks and normal traffic by the Artificial Neural Networks increased up to 17.6% using the Genetic Algorithm. The Fuzzy System presented a slight gain of about 5% on hit rate of attacks and normal traffic. However, when Neural Networks and Fuzzy System were combined, forming the Neuro-Fuzzy System, we have obtained a gain around 30% on hit rate, when compared to the original work. Then, we present some conclusions of this work and some possible future work.