Given a model with multiple input parameters, and multiple possible sources for collecting data for those parameters, a data collection strategy is a way of deciding from which sources to sample data, in order to reduce the variance on the output of the model. Cain and Van Moorsel have previously formulated the problem of optimal data collection strategy, when each parameter can be associated with a prior normal distribution, and when sampling is associated with a cost. In this paper, we present ADaCS, a new tool built as an extension of PRISM, which automatically analyses all possible data collection strategies for a model, and selects the optimal one. We illustrate ADaCS on attack trees, which are a structured approach to analyse the impact and the likelihood of success of attacks and defenses on computer and socio-technical systems. Furthermore, we introduce a new strategy exploration heuristic that significantly improves on a brute force approach.
CITATION STYLE
Mace, J. C., Thekkummal, N., Morisset, C., & Van Moorsel, A. (2017). ADaCS: A tool for analysing data collection strategies. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 10497 LNCS, pp. 230–245). Springer Verlag. https://doi.org/10.1007/978-3-319-66583-2_15
Mendeley helps you to discover research relevant for your work.