Ph.D. / This thesis is a theoretical treatise on a proposed new computer security system, based on a biological immune system. Modem day network-centric computing is fast approaching the density and complexity of biological organisms, making biological and computer analogies relevant and meaningful. The success of biological immune systems in protecting life over countless millennia is well known. It is therefore postulated that a highly effective defensive mechanism can be developed, to transparently enforce an acceptable level of security in very extensive and complex computer networks and systems. It can be done by building very basic, but specialized autonomous software agents, functioning in a hierarchical system, that follow basic rules that can be deduced from biological immune systems. The computer security model proposed in this thesis does not require radical new technologies and it is extremely simple. The complexity however, lies in the effective implementation of the model. Three building blocks support the computer immune model: Biology, software agent technology and nanotechnology. The main features of the model are: • Firstly, it allows computer systems to automatically and transparently protect themselves, by using mobile autonomous intelligent software agents in an artificial immune system, based on biological immune systems. • Secondly, it allows computer systems to be pro-active in protecting themselves by being able to discern between which components are part of themselves (trusted system), and which components are foreign and may thus be harmful. Although part of the model relies on specialist human collaboration and international security standards, the main thrust is the heuristic ability of the proposed computer immune system. It allows systems to learn to recognize and cope with previously unknown cyber-antigens, automatically share the information amongst the participating computer systems, and thus 'inoculate' them to render them immune to similar attacks.

Made available in DSpace on 2016-12-23T14:33:37Z (GMT). No. of bitstreams: 1 dissertacao.pdf: 1052399 bytes, checksum: b17ce224ca3822e277c997fd00bd2c67 (MD5) Previous issue date: 2006-11-29 / Este trabalho apresenta um Sistema Imune Artificial (SIA) para tratar problemas de escalonamento. O Sistema Imunológico Artificial desenvolvido neste projeto baseia-se na estrutura arquitetura e funcionamento dos Sistemas Imunes Biológicos ou Naturais. O uso de Algoritmo Genético (AG) fez-se necessário para gerar os indivíduos a serem escalonados, representando os antígenos e anticorpos do SIA. Cada indivíduo gerado pelo AG representa um conjunto de tarefas processadas em um conjunto de máquinas. Os indivíduos são avaliados por uma função de aptidão que representa o processo de seleção natural. A evolução dos indivíduos e consequentemente das populações são obtidas aplicando-se os operadores genéticos de crossover e mutação. As tarefas e as máquinas, utilizadas para o escalonamento, representa o problema de Job Shop Scheduling (JSS). Ao problema, foram aplicados alguns testes clássicos da literatura, onde se verificou a viabilidade dos SIA para tratamento de problemas de escalonamento. Ainda com os testes, pode-se observar o comportamento do sistema durante toda a execução, possibilitando assim, uma análise criteriosa das funcionalidades do sistema e dos resultados gerados pela massa de teste, observados durante um período de tempo. A representação dos sistemas imunológicos naturais através de algoritmos computacionais tem inspirado pesquisadores de todo o mundo, a motivação é que os sistemas imunológicos possuem características de paralelismo adaptabilidade e aprendizagem, além da possibilidade de serem aplicados em diversos problemas das mais diversas áreas, devido sua portabilid ade. / This work presents an Artificial Immune System (AIS) to deal with problems scheduling. The Artificial Immunologic System developed in this project was based on the structure, architecture and functioning of the Biological or Natural Immune Systems. The use of Genetic Algorithm (GA) became necessary to represent the antibodies and antigens of the AIS. Each individual generated for the GA represented a processed task set library in a set of machines. The evaluation of each individual was given by a fitness function that represents the process of natural selection. The evolution of the individuals, and population as a consequence was obtained by applying the genetic operators of crossover e mutation. The machines and the tasks used for the scheduling represent the problem of Job Shop Scheduling (JSS). Some classic tests of the literature where applied to the problem in order to verify the viability of the AIS on the treatment of task of scheduling problems. Those tests also demonstrated the system s behavior its entire execution, therefore, allowing for a detailed analysis of the system s functionalities sets for certain time period. The representation of the natural immunologic systems through computational algorithms inspires from all over world researchers. The motivation is that the immunologic systems possess parallelism characteristics adaptability and learning, which can be applied in several problems found in many areas, had its portability.

Made available in DSpace on 2016-12-23T13:44:55Z (GMT). No. of bitstreams: 1 dissertacao.pdf: 1052399 bytes, checksum: b17ce224ca3822e277c997fd00bd2c67 (MD5) Previous issue date: 2007-03-23 / This work presents an Artificial Immune System (AIS) to deal with problems scheduling. The Artificial Immunologic System developed in this project was based on the structure,architecture and functioning of the Biological or Natural Immune Systems. The use of Genetic Algorithm (GA) became necessary to represent the antibodies and antigens of the AIS. Each individual generated for the GA represented a processed task set library in a set of machines. The evaluation of each individual was given by a fitness function that represents the process of natural selection. The evolution of the individuals, and population as a consequence was obtained by applying the genetic operators of crossover e mutation. The machines and the tasks used for the scheduling represent the problem of Job Shop Scheduling (JSS). Some classic tests of the literature where applied to the problem in order to verify the viability of the AIS on the treatment of task of scheduling problems. Those tests also demonstrated the system s behavior its entire execution, therefore, allowing for a detailed analysis of the system s functionalities sets for certain time period. The representation of the natural immunologic systems through computational algorithms inspires from all over world researchers. The motivation is that the immunologic systems possess parallelism characteristics adaptability and learning, which can be applied in several problems found in many areas, had its portability. / Este trabalho apresenta um Sistema Imune Artificial (SIA) para tratar problemas de escalonamento. O Sistema Imunológico Artificial desenvolvido neste projeto baseia-se na estrutura arquitetura e funcionamento dos Sistemas Imunes Biológicos ou Naturais. O uso de Algoritmo Genético (AG) fez-se necessário para gerar os indivíduos a serem escalonados, representando os antígenos e anticorpos do SIA. Cada indivíduo gerado pelo AG representa um conjunto de tarefas processadas em um conjunto de máquinas. Os indivíduos são avaliados por uma função de aptidão que representa o processo de seleção natural. A evolução dos indivíduos e consequentemente das populações são obtidas aplicando-se os operadores genéticos de crossover e mutação. As tarefas e as máquinas, utilizadas para o escalonamento, representa o problema de Job Shop Scheduling (JSS). Ao problema, foram aplicados alguns testes clássicos da literatura, onde se verificou a viabilidade dos SIA para tratamento de problemas de escalonamento. Ainda com os testes, pode-se observar o comportamento do sistema durante toda a execução, possibilitando assim, uma análise criteriosa das funcionalidades do sistema e dos resultados gerados pela massa de teste, observados durante um período de tempo. A representação dos sistemas imunológicos naturais através de algoritmos computacionais tem inspirado pesquisadores de todo o mundo, a motivação é que os sistemas imunológicos possuem características de paralelismo adaptabilidade e aprendizagem, além da possibilidade de serem aplicados em diversos problemas das mais diversas áreas, devido sua portabilidade.