 Hi, I'm Antonijon and presenting privacy protected map using adaptive kernel density estimation. So last year I was able to present SST Spatial with the user 2019 about SST Spatial and this slide is a recap of that. So we have added some more methods which I'm going to talk about more. So SST Spatial is a package to create density maps that protect privacy of individual observations the several methods, for example, for creating master maps, pledging density, value density, and mean density, these methods for finding out which locations are sensitive, so reveal privacy protected privacy and sensitive information, so it's a plotting method and has a sensitive method. And of course it includes some methods for protecting data, so protect smooth by smoothing data, protect plot tree, and removing sensitive locations at all from your map. What we have added is adaptive smoothing with protect adaptive. So for example if you look at the dwellings data which is included in the package, you can see that it contains locations which are realistic, but also two simulated variables consumption and unemployedness. If you create a SST Raster object, you will see that 42% of this particular example is sensitive, so it runs the risk of revealing the privacy of these consumption. If you plot this with plot function, you can see that the consumption pattern is over on the left side and on the right side you can see which part of the data is sensitive, considered sensitive, which is indicated with red. So how can you reduce the sensitive team, so you can use a course grid, so more course grade, remove a sensitive location at all, you can aggregate sensitive cells hierarchically, you can apply spatial smoothing with project smooth, which is a very fine method, and you can apply adaptive smoothing, which we have added to the package. So why adaptive smoothing? Well, special phenomena are not uniformly distributed, so it's a bit easy to think that one bandwidth will take care of everything, it's not optimal, and in practice your choice of bandwidth is a trade-off between more detailed areas and very less detailed areas, populated areas, and adaptive smoothing has problems to take this density into account, so high populated areas will be detailed, low populated areas are more smooth. So we use some math to define what is sensitive and how to protect it. These are implemented in the package itself, so percent rule and the method works by adding adaptive noise, so areas which are less populated contain more noise, and high populated areas contain less noise. The deputation is from the covariance matrix, so in words the noise adds depths to intensity and failure density and low density areas generate more noise, which is fine, but can also give some potential artifacts because the noise can be very big. But for real use cases this seems to work quite well, so this is one example of this real use case, this is from the energy pattern from enterprises, so special energy pattern. On the left side you can see just normal smoothing and on the right side you can see adaptive smoothing, which is very similar and you can see that the difference are due to these areas where more sensitive, so this looks quite nice. So if you're interested in creating privacy protected maps, this special would be fine for you I guess, just five lines you can create the privacy protected map, I create an SSE arrest object, applying smoothing, remove sensitive cells, remaining sensitive cells, extract the raster and plot remaining raster. So thank you for your attention, if you have any questions or just curious, just install the package, it's in spatial and feedback and suggestions are most welcome. Thank you for your attention.