Exploring the solution space
In this perspective, the solution space can be browsed and transformations can be applied to the input data set, as is shown in the following screen shot.
In area 1 a subset of the current solution space is displayed. The transformations which are shown here, can be selected by using filters that are provided in area 2. Area 3 implements a clip board in which transformations can be stored and annotated, while area 4 displays properties of the currently selected transformation.
Apart from some shortcuts to menu entries, ARX's application toolbar also displays some information from this perspective: (1) the currently selected transformation, (2) the currently applied transformation as well as some statistics regarding the solution space. These statistics include the number of privacy-preserving transformations and the total number of transformations in the solution space as well as some statistics about the previous search process.
Each transformation is identified by the generalization levels which it specifies for the quasi-identifiers in the input data set.
Visualization of the solution space
In the center of the screen, the view displays a subset of the current solution space. The space may be visualized in different ways. First, it can be displayed as a Hasse diagram of the generalization lattice.
Here, each node represents a single transformation, which is identified by the generalization levels that it specifies for the quasi-identifiers in the input data set. Transformations are characterized by three different background colors:
- Green denotes that the transformation results in a privacy-preserving data set.
- Red denotes that the transformation does not result in an privacy-preserving data set.
- Orange denotes that the transformation is the optimal solution according to the specified utility measure.
Transformations that have not been pruned during the de-identification process are marked with a thick blue border. Dragging the mouse while holding the left button will move the current section of the solution space. Dragging the mouse while holding the right button will zoom in and out of the current view. Hovering the mouse over a transformation will display a tool tip with basic information about the transformation. Clicking a transformation with the left button will select it, while clicking a transformation with the right button will bring up a context menu. Here, transformations can either be applied to the current input data set or added to the clipboard.
The solution space may also be visualized as a list.
Here, each transformation is represented by one row. Rows are sorted by score. Data utility and the anonymity property are highlighted graphically. Thirdly, the solution space may be displayed as a set of tiles.
This view is very similar to the list described in the previous paragraph. The color and border of an element encodes information about the anonymity property, while the background color represents data utility. Elements are also ordered by their utility.
Filtering the solution space
The screen shot shows controls for filtering the subset of the solution space that will be displayed in this perspective. First, a quasi-identifier can be selected and transformations can be restricted to ones defining the selected generalization levels for the selected quasi identifier. Second, the displayed transformations can be restricted to transformations that do or do not preserve privacy according to the specified privacy model.
The sliders at the bottom allow to display only transformations with a score that is in between the configured bounds. Note that the scores of transformations which have not been checked explicitly are not known and that they are therefore estimated with bounding intervals that are computed from other solutions in the search space. These estimates might be wrong when a heuristic de-identification process has been performed.
The screen shot shows the clip board available for organizing and annotating transformations. Right-clicking an entry opens up a context menu, which allows to remove the transformation, apply it to the current data set and to edit its annotations. The buttons in the upper right area support removing all entries, changing their order and sorting them according to the resulting reductions in data utility.
Properties of transformations
Basic information about the currently selected transformation will be displayed in the this area. The highest and lowest scores are shown as absolute values and as values relative to the highest and lowest scores of all transformations in the solution space.
The view also displays the number of successors and predecessors of a transformation, the generalization levels which are specified for the quasi-identifiers and whether the transformation has been pruned during the de-identification process.