The works presented in this dissertation are representative of my activities in the field of resilience and privacy for ubiquitous mobile systems. They are organized according to three research areas : the use of reflection for the construction of dependable architectures, resilient mobiquitous systems (architectures, algorithms and evaluation) and geo-privacy. The first research area concerns fault tolerance for distributed systems, from an architectural and language viewpoint. In this work, I have investigated the use of compile-time reflection to facilitate the implementation of fault tolerance mechanisms independently of the application. The use of reflection has been studied in the context of both theoretical work on multi-level reflexivity, and more practical work on the implementation of reflective off-the-shelf components in embedded software architectures. The second research area concerns fault tolerance in mobile systems. My angle of attack was to consider mobility as an asset and not as a potential difficulty. This approach led me to explore the concept of geographical communication groups : how to define a group of communicating entities according to their respective location or as a function of their proximity. Then, I proposed a cooperative backup service, where the participating nodes offer a p2p secure storage service that they can use to back up their critical data. This approach was also followed to provide a virtual black box service for cars. These works were treated from both algorithmic and architectural viewpoints, and also in terms of analytical and experimental dependability assessment. As part of my research on the resilience of mobiquitous systems, ethical issues were raised : how to exploit mobility data of individuals, while preserving their privacy ? This raised my interest in what can be called geo-privacy. This third area is now the major focus of my research, both in terms of attacks and protection. We propose a Markov mobility model as a compact, accurate, understandable and easily adaptable tool to represent the mobility of an individual. Based on this mobility model, we propose several attacks that target, e.g., prediction of future mobility, and de-anonymisation. Regarding the protection of geo-privacy, we are currently working on middleware-level abstractions, such as locanyms and location proofs to provide a secure and private architecture for location-based systems. These works were linked with each other through a mixture of chance and serendipity, and the pursuit of a common goal : providing means for architectural and algorithmic resilience of current computer systems, namely distributed, mobile, ubiquitous systems. Many avenues of research are still open and are discussed.
