Uncertainty and Provenance in Collaborative Situation Awareness

The so-called big data revolution has been characterised by an increase in sources of data as well as in the volume of data to be processed.

In many cases—for example, network behaviour and control, security monitoring, enterprise management information—the data for situation awareness and decision-making is drawn from multiple sources and must be integrated into a coherent whole as far as possible.

This process generally requires both machines and human analysts and experts, It includes compensating for different formats, different granularities and resolutions, identifying and correcting errors (both systematic and intermittent), as well as managing uncertainties and gaps in data. Often the process requires assumptions and choices to be made in arriving at a reasonably robust overview of a situation—for example, in deciding that a failed attempt to access a building is potentially malicious, we might need to take account of someone's recent travel, long term patterns of behaviour, current schedules of close colleagues, etc. where each of these components may have been derived from lower-level raw data. Provenance in this context refers to the derivation pathways and their overall reliability.

In this talk, I will describe the use of graded (fuzzy) representations in modelling and managing the uncertainties, reliability and granularity of derived data in combining sources for situation awareness.