09 février 2018
Auditoire BARB94 - Place Sainte Barbe, 1
Software Requirements Engineering: A Risk-Driven Approach
Requirements Engineering (RE) is concerned with t he elicitation, evaluation, specification, analysis and evolution of the requirements on a software-intensive system. Risk analysis at RE time aims at an ticipating adverse conditions preventing the system from ach ieving its mission. In goal-oriented RE, a goal model shows how the system's objectives contribute to each other. Obstacle analysis is a goal-oriented form of risk analysis aimed at increasing re quirements completeness. An obstacle to a goal is a precondition for the non-satisfaction of this goal. An obstacle model shows how ob stacles contribute to each other in preventing goals from being satisfied. Obstacle analysis iterates on the identification of ob stacles, the assessment of their likelihood and criticality, and the con trol of likely and critical obstacles.
The thesis presents a quantitative framework for assessing and con trolling obstacles to probabilistic goals. The latter must be satisfied in a specified percentage of cases at least. In this fra mework, domain experts estimate the likelihood of fine-grained ob stacles together with their uncertainty margins. These estimates are up-propagated through the obstacle and goal models in order to quantitatively determine the likelihood of obstacles and the sev erity of their consequences. Comparing the computed satisfaction rate of high-level goals in the model with their required satisfaction rate yields measures of obstacle criticality. Countermeasures to most likely and critical obstacles are then identified. Those maximizing the satisfaction rate of high-level goals while minimizing their cost are selected for integration into the goal model. Our techniques are extended to support runtime system self-adaptation towards better satisfaction of high-level goals.
Membres du jury:
Prof. Axel van Lamsweerde (UCL), promoteur
Prof. Peter Van Roy (UCL), président
Prof. Charles Pecheur (UCL), secrétaire
Prof. Jeff Kramer (Imperial College London, UK)
Prof. Carlo Ghezzi (Politecnico di Milano, Italy)
Prof. Emmanuel Letier (UCL London, UK)