Conventional wisdom suggests that in order to build a secure system, security must be an integral component in the system design. However, cost considerations drive most system designers to channel their efforts on the system's performance, scalability and usability. With little or no emphasis on security, such systems are vulnerable to a wide range of attacks that can potentially compromise confidentiality, integrity and availability of sensitive data. It is often cumbersome to redesign and implement massive systems with security as one of the primary design goals. This thesis advocates a proactive approach that cleanly retrofits security solutions into existing system architectures. The first step in this approach is to identify security threats, vulnerabilities and potential attacks on a system or an application. The second step is to develop security tools in the form of customizable and configurable plug-ins that address these security issues and minimally modify existing system code, while preserving its performance and scalability metrics. This thesis uses overlay network applications to shepherd through and address challenges involved in supporting security in large scale distributed systems. In particular, the focus is on two popular applications: publish/subscribe networks and VoIP networks. Our work on VoIP networks has for the first time identified and formalized caller identification attacks on VoIP networks. We have identified two attacks: a triangulation based timing attack on the VoIP network's route set up protocol and a flow analysis attack on the VoIP network's voice session protocol. These attacks allow an external observer (adversary) to uniquely (nearly) identify the true caller (and receiver) with high probability. Our work on the publish/subscribe networks has resulted in the development of an unified framework for handling event confidentiality, integrity, access control and DoS attacks, while incurring small overhead on the system. We have proposed a key isomorphism paradigm to preserve the confidentiality of events on publish/subscribe networks while permitting scalable content-based matching and routing. Our work on overlay network security has resulted in a novel information hiding technique on overlay networks. Our solution represents the first attempt to transparently hide the location of data items on an overlay network.

Orientador : Ricardo Dahab. / Tese (doutorado) - Universidade Estadual de Campinas, Instituto de Computação / Made available in DSpace on 2018-08-21T17:02:27Z (GMT). No. of bitstreams: 1 GalloFilho_RobertoAlves_D.pdf: 5999506 bytes, checksum: 6ef66e76246dddb7de30593abff60bc5 (MD5) Previous issue date: 2012 / Resumo A concepção de sistemas seguros demanda tratamento holístico, global. A razão é que a mera composição de componentes individualmente seguros não garante a segurança do conjunto resultante2. Enquanto isso, a complexidade dos sistemas de informação cresce vigorosamente, dentre outros, no que se diz respeito: i) ao número de componentes constituintes; ii) ao número de interações com outros sistemas; e iii) 'a diversidade de natureza dos componentes. Este crescimento constante da complexidade demanda um domínio de conhecimento ao mesmo tempo multidisciplinar e profundo, cada vez mais difícil de ser coordenado em uma única visão global, seja por um indivíduo, seja por uma equipe de desenvolvimento. Nesta tese propomos um framework para a concepção, desenvolvimento e deployment de sistemas baseados em hardware que é fundamentado em uma visão única e global de segurança. Tal visão cobre um espectro abrangente de requisitos, desde a integridade física dos dispositivos até a verificação, pelo usuário final, de que seu sistema está logicamente íntegro. Para alcançar este objetivo, apresentamos nesta tese o seguinte conjunto de componentes para o nosso framework: i) um conjunto de considerações para a construção de modelos de ataques que capturem a natureza particular dos adversários de sistemas seguros reais, principalmente daqueles baseados em hardware; ii) um arcabouço teórico com conceitos e definições importantes e úteis na construção de sistemas seguros baseados em hardware; iii) um conjunto de padrões (patterns) de componentes e arquiteturas de sistemas seguros baseados em hardware; iv) um modelo teórico, lógico-probabilístico, para avaliação do nível de segurança das arquiteturas e implementações; e v) a aplicação dos elementos do framework na implementação de sistemas de produção, com estudos de casos muito significativos3. Os resultados relacionados a estes componentes estão apresentados nesta tese na forma de coletânea de artigos. 2 Técnicas "greedy" não fornecem necessariamente os resultados ótimos. Mais, a presença de componentes seguros não é nem fundamental. 3 Em termos de impacto social, econômico ou estratégico / Abstract: The conception of secure systems requires a global, holistic, approach. The reason is that the mere composition of individually secure components does not necessarily imply in the security of the resulting system4. Meanwhile, the complexity of information systems has grown vigorously in several dimensions as: i) the number of components, ii) the number of interactions with other components, iii) the diversity in the nature of the components. This continuous growth of complexity requires from designers a deep and broad multidisciplinary knowledge, which is becoming increasingly difficult to be coordinated and attained either by individuals or even teams. In this thesis we propose a framework for the conception, development, and deployment of secure hardware-based systems that is rooted on a unified and global security vision. Such a vision encompasses a broad spectrum of requirements, from device physical integrity to the device logical integrity verification by humans. In order to attain this objective we present in this thesis the following set of components of our framework: i) a set of considerations for the development of threat models that captures the particular nature of adversaries of real secure systems based on hardware; ii) a set of theoretical concepts and definitions useful in the design of secure hardware-based systems; iii) a set of design patterns of components and architectures for secure systems; iv) a logical-probabilistic theoretical model for security evaluation of system architectures and implementations; and v) the application of the elements of our framework in production systems with highly relevant study cases. Our results related to these components are presented in this thesis as a series of papers which have been published or submitted for publication. 4Greedy techniques do not inevitably yield optimal results. More than that, the usage of secure components is not even required / Doutorado / Ciência da Computação / Doutor em Ciência da Computação