Obfuscating event location on map

Question

I'm creating an events app where users can host events at their place and invite people. I want the list of hosted events to be publicly viewable on a map, designated by some icon. BUT, public event locations will be obfuscated by +1/-1 miles for both latitude and longitude. If you get invited to an event, you'll receive the actual address. This is to protect user's home address.

Seeing as how markers are used to indicate exact location, they seem inappropriate for obfuscated locations. I want users to get a feel of the general area of the event. I was thinking some sort of half transparent, circle figure, whose center is on the obfuscated lat/lng position.

What would be the best shape/icon for representing an "area" on the map?

Answer 1

I would not use the map as a user interface when the address is still private information. Because, the map is used to represent precise locations at some level you will always run into the issue of privacy.

The map is useful if the user gets to choose events they would like to attend based on location. So I question the usefulness of a map of events if users have to be invited to be given the location for the event. I would use a different UI to list the events and use the map once the user has been invited. You can filter the list to include only those that fit within a particular radius from the user's location. To preserve privacy you can use buckets instead of specifying the actual radius to the user, for example, nearby, a bit out of the way and far far away.

The options you gave for representing an area are all fine but keep in mind that they don't ensure privacy at the level of the area you choose to show, for example, neighbourhood or town.

Answer 2

If your mapping framework provides the feature, highlighting a "suburb" might be appropriate for your use case. Of course the actual size of what is considered a suburb varies widely from region to region.

This screenshot shows google maps highlighting the suburb of Tanunda, South Australia.

Answer 3

Another option, not sure if you've considered this: Ask the person setting up the event for a nearby public landmark (library, shopping center, etc.). Then use that location. That saves you from needing a lot of local knowledge and should handle a pretty wide variety of population densities.

It also nicely handles things like rivers. Around here, for example, the Potomac river divides Maryland and Virginia. A location on the Virginia side can be 10 minutes away, but move it 1,000 feet across the river, and suddenly it's an hour drive away. Your users will know this and pick a landmark on the correct side of the river.

Answer 4

Your idea sounds like a reasonable solution.

I would, however, ensure that this offset is not just randomly generated every time the map loads... With enough randomly-generated offsets, it would be possible to derive the actual center of this distribution.

Additionally, if you'd like to really emphasize that it's a general area, don't display a defined border to the circle, but rather just a blurred circular overlay that represents the general area. Something like the following, perhaps?

_{Note: It's just an illustration. The pattern is terrible, I know, and I agree--it was a preset option and I didn't bother adjusting it. I figured it was good enough for a mockup...}

Answer 5

Use the postal codes and corresponding area locations. It's an already implemented obfuscation/aggregation system adjusted for population density.

Answer 6

You obviously cannot just take your precise loction and cover it with a blurb of grey; it still pinpoints right to the destination.

What you can (easily) do, though, is pick the next city, point out that city, and then make it very clear from the UI that the location is not by chance the center of that city. Make your location maker so obviously in the geographic center of the city you picked so that nobody would get the idea that the location is actually in the inner city, and label it something like "located somewhere near CityX" or whatever good idea you come up with.

To locate a city, make a query where you ask for "between X and Y km/miles away from the actual location + at least Z residents". Z does not need to be huge, but should likely be something like 3000 or up, to rule out small countryside villages.

If you get no results (i.e., for a very quiet rural area), then you need to scale up; i.e., make it "somewhere in state X" or something like that.

The information content for the user is mediocre, and like other answers said you could just skip the map display completely, but this is a solution to your question which will make your stakeholders happy and is a snazzy feature unless your clients happen to live primarily in far off farm villages.

Answer 7

If you limit the zoom and use a fairly large blob rather than a pin marker, this has a natural feel of low precision. Here's a mockup:

Only major roads are shown but you can still see which side of the city the event is; you can't navigate.

I know some mapping APIs allow custom markers; I believe some allow limited zoom, at least when used embedded. The location fed to the API should be of limited precision (i.e. rounded).

If all else fails, you can do just this but display an image. I know of one site that does this quite successfully to allow people to decide whether an event is in the right place for them -- it's a bit oldschool but it works

Answer 8

If you internally divide the world space into a grid (say, 1x1 mile squares), then events can be classified as being inside some single square and whenever you render the map, you could just randomly place the marker for that event anywhere within that square. That way, you're not giving any hint about center and you still get discrete markers so it'll be easy to tell relative density.

You could even assign colors to grid segments based on number of events, skipping the markers so it's more like a heatmap, or have a single marker on each populated square representing roughly how many are in that area. And if you wanted to calculate distance to events you could assume each event's position is the center of the square.

Answer 9

How about displaying an event's location as "2 miles north of you" or something like that? In some areas, neighborhoods are named, so you could say: "4 miles northwest in the Lake Highlands area." For something close enough to risk a potential giveaway, you could say: "Just down the road."

Answer 10

You could internally divide your map into predefined hexes with each hex being maybe 1 mile/2 km across. For any given event, its location will fall into exactly one hex. You could then show a circle on your map with no defined edge (either a shaded area with a faded edge, or just a shaded area with no outline) that encircles the hex -- the event will be guaranteed to be within that circle, but users will not be able to triangulate where an event occurs within that circle.

You could then count the user as being in their own hex, and provide the approximate distance based on hex-to-hex distance, e.g. "this event is about 5 miles away". This will provide a ±1mi/2km accuracy (if the user and event are on far corners of their respective hexes it'd be just under 6 miles, and if they're on the near corners it'd be just over 4 miles) but users (or bots) will still not be able to triangulate an event's location beyond the 1mi/2km hex.

If it's a really rural area where the location is literally the only place in a 1mi+ radius then yes, it could be deduced, but this is an extreme edge case.

The circles that the users see will effectively have overlap, but because the underlying "architecture" is based on non-overlapping hexes, you will never have a case where a particular location shows up in more than one circle.

Answer 11

Instead of showing the location of each event, show only the average-locations of groups of 2 or more events. Here's a quick mock-up:

_{The pink dots are because I made this from a screenshot of flickr.com/map.  IIRC, Flickr used to have the UI I'm describing— images were lumped together into “aggregate dots” — but sadly that implementation of the map interface seems to be gone.}

Your algorithm should never display a single “(1)” dot on the map— if there's only one event in a location, it should look further out until it finds another event and then average the locations together— e.g. in the rare edge case that a metro area has only 2 events on the opposite sides of the downtown area would show a “(2)” dot right in the middle of the downtown area.  The more events there are in an area, the more accurate the location effectively becomes, but there should be a hard limit to the accuracy too— events within a minimum distance (e.g. 1000 meters) are always grouped together, and so you end up with higher-numbered bubbles as seen above.

In terms of privacy, the lat/long accuracy is tied to the density of events… and high event densities are almost certainly only going to occur where there's also high population/venue densities.

The only edge-case I can think of in which an event's location could become unobfuscated is the situation that multiple in-app events are added for the same real-world event.  Doing a filter on your data before mapping it that removes duplicate addresses could be one way to solve this.

Answer 12

Use a fixed size grid for your map and chosse the grid size according to your need (e.g. 2mile squares). Then highlight the grid containing the event location.
Note that the grid shoud have a fixed start point.