Abstract

Research: Privacy types revisited, Sören Preibusch

Privacy types structure a population of consumers into several groups that exhibit similar concerns about revealing personal information. Originally developed for an offline world, privacy types such as “fundamentalists” or “unconcerned” have been applied subsequently to online populations to characterise Web users with pronounced or non-existent concerns about data protection respectively. The main promise of privacy types is an intuitive, easy-to-use taxonomy of privacy concerns to support academic research and corporate planning alike. This work-in-progress talk suggests the assumptions for establishing privacy types are not necessarily met. Based on empirical data collected in a winter 2009 field experiment, the speaker argues that prototypical privacy preferences are hard to discern. Even fine-grained clustering achieves poor coverage of the entire online population. Alternative approaches to this apparent heterogeneity in privacy preferences are discussed along with managerial implications.

Research: Predicting the Performance of Virtual Machine Migration, Sherif Akoush

Live migration is a particularly useful feature in virtualised infrastructure. However, in order to be useful on a large scale it is important to be able to predict migration times accurately. In this presentation I characterise the parameters affecting live migration in Xen and show that increasing link speeds can provide significant gains for some applications. I further highlight significant variations in migration times and provide two simulation models that are able to predict virtual machine migration times to within more than 90% accuracy for both synthetic and real-world benchmarks.