This paper proposes an approach for automatic, service-driven configuration of networked IT systems focused on achieving a specific set of dependability properties. An automatic process starts from a service-level model to generate dependable configurations for the managed IT infrastructure. The process uses an ontology to model the services and their dependability requirements, the IT infrastructure, the available dependability mechanisms, and their configuration. Configurations are computed by model transformation rules which implement various dependability strategies with different degrees of requirements' satisfaction. Alternative configurations are generated to allow reconfiguring the system taking into account incidents or other operational conditions. A sample system hosting services based on web technologies is used as a proof-of-concept to illustrate application and extent of our approach.

Denial-of-Service attacks, and jamming in particular, are a threat to wireless networks because they are at the same time easy to mount and difficult to detect and stop. We propose a distributed intrusion detection system in which each node monitors the traffic flow on the network and collects relevant statistics about it. By combining each node’s view we are able to tell if (and which type of) an attack happened or if the channel is just saturated. However, this system opens the possibility for misuse. We discuss the impact of the misuse on the system and the best strategies for each actor.

Widespread adoption of wireless devices has considerably enlarged the domain of user applications, but consequently specific issues related to a secure use of these devices arise, in particular when user privacy may be at stake. This work aims at assuring an adequate degree of privacy when using pervasive wireless devices, by means of a modified flooding routing algorithm and known anonymity techniques, which however have not yet been applied to wireless scenarios. The level of privacy reached is evaluated by traffic matrices and experimental tests.

Emerging ubiquitous computing leverages the need for automated trust management models. We take a domestic network of both fixed and mobile nodes as our reference scenario and investigate how nodes can build trust exclusively based on their past experience. In a home network, trust propagation can be superfluous, while direct evaluation of peers' behaviour during repeated interactions has a key role. We present experimental results on the use of statistics for automated trust building and outline numerous issues that should be analysed further.