Sequential binning privacy model for data with prior knowledge of feature dependencies

Unlike physiological measurements taken in conventional medical environments, the medical Internet of Things (MIoT) is likely to result in duplicate and overlapping measurements, which can be associated with different personally identifiable items of information. Moreover, it can be expected that the ensemble of MIoT measurements may change as devices are added and removed.
This poses new challenges for modelling privacy and to optimise anonymisation. We propose to extend differential privacy models to explicitly incorporate feature dependencies, assuming that additional (external) knowledge of these relations and models can be represented in the form of joint probability distributions, such as mutual information. We propose an enhanced definition of differential privacy in conjunction with a realisation for non-randomising anonymization strategies such as binning, reducing the extent of binning required and preserving more valuable information for researchers.
As part of this effort, we also propose a model for feature collection and addition in the form of partial orders to capture the aforementioned dynamic effects. This allows the formulation of privacy conditions over the evolving set of features such that each feature can be associated its own allowance for additional information either based on a priori information about a sensor, or on external knowledge on dependencies and feature probabilities.