Applications that continuously gather and disclose personal information about users are increasingly common. While disclosing this information may be essential for these applications to function, it may also raise privacy concerns. Partly, this is due to frequently changing context that introduces new privacy threats, and makes it difficult to continuously satisfy privacy requirements. To address this problem, applications may need to adapt in order to manage changing privacy concerns. Thus, we propose a framework that exploits the notion of privacy awareness requirements to identify runtime privacy properties to satisfy. These properties are used to support disclosure decision making by applications. Our evaluations suggest that applications that fail to satisfy privacy awareness requirements cannot regulate users' information disclosure. We also observe that the satisfaction of privacy awareness requirements is useful to users aiming to minimise exposure to privacy threats, and to users aiming to maximise functional benefits amidst increasing threat severity.

We propose the use of forensic requirements to drive the automation of a digital forensics process. We augment traditional reactive digital forensics processes with proactive evidence collection and analysis activities, and provide immediate investigative suggestions before an investigation starts. These activities adapt depending on suspicious events, which in turn might require the collection and analysis of additional evidence. The reactive activities of a traditional digital forensics process are also adapted depending on the investigation findings.

In a dynamic environment where context changes frequently, users’ privacy requirements can also change. To satisfy such changing requirements, there is a need for continuous analysis to discover new threats and possible mitigation actions. A frequently changing context can also blur the boundary between public and personal space, making it difficult for users to discover and mitigate emerging privacy threats. This challenge necessitates some degree of self-adaptive privacy management in software applications.

This paper presents SecuriTAS, a tool to engineer adaptive security. It allows software designers to model security concerns together with the requirements of a system. This model is then used at runtime to analyze changes in security concerns and select the best set of security controls necessary to protect the system.

This paper presents a novel approach to the design of self-adaptive service-oriented applications based on a new model called service tiles. The approach allows designers to develop a service-oriented system by building an assembly of component services that accomplishes the given goal. The assembly is computed automatically starting from the specification of a subset of the whole system, a few constraints, and the goals the application should fulfill. An application designed according to the service-tile model can also dynamically self-adapt by replacing, in part or entirely, services in the assembly whenever they fail or the application context changes. The service-tile design technique has been implemented in a prototype and some experiments with several examples demonstrate the feasibility of the approach and its practical efficiency.

An important problem that arises during the execution of service-based applications concerns the ability to determine whether a running service can be substituted with one with a different interface, for example if the former is no longer available. Standard Bounded Model Checking techniques can be used to perform this check, but they must be able to provide answers very quickly, to avoid that the check may affect the operativeness of the application, instead of aiding it. The problem becomes even more complex when conversational services are considered, i.e., services that expose operations that have Input/Output data dependencies among them. In this paper we introduce a formal verification technique for an extension of Linear Temporal Logic that allows users to include in formulae constraints on integer variables. This technique applied to the substitutability problem for conversational services is shown to be considerably faster and with smaller memory footprint than existing ones.