potential security issue

[ligi]

Really like the idea of the project!

But just wondering: aren't you leaking so much metadata to https://map-api-direct.foam.space and your tileprovider that your keyspace is very small and when you know this metadata your key can easily be brute forced?

[tschubotz]

Hey @ligi, Absolutely, potentially leaked metadata would make it easier for brute force attacks. That would have to be prevented for sure. Currently, the biggest issue is the keyspace anyway as FOAM only has ~10k POI on their platform. So even without additional knowledge via meta data, it is not secure enough :)

[ademidun]

Agreed with @ligi this is super cool!

I think you could solve the FOAM 10k limitation by using an aggregate of different geocoding apis and taking the average of the lat_long, then using the same geohashing algorighm you already use

I know the use of centralized providers may not be ideal, but decentralized, centralization could be thought of as the best of both worlds. I'm pretty sure the Chainlink Pricing Oracle works similarly [pricing_oracle].

• "Berlin" -> [google_api] -> berlin_lat_long_1
• "Berlin" -> [open_weather_api] -> berlin_lat_long_2
• "Berlin" -> [mapquest_api] -> berlin_lat_long_3
• "Berlin" -> [bing_mapbox_api] -> berlin_lat_long_4

berlin_lat_long_avg = avg(berlin_lat_long_1, berlin_lat_long_2, berlin_lat_long_3, berlin_lat_long_4)

berlin_lat_long_avg -> [geohash] -> privatekey_1_of_5

This allows the keyspace to be infinitely large as users can enter locations at any desired granularity

E.g. I could enter a location to be any choice of: Germany, Berlin, Kreuzberg, or Görli[gorli]

Caveats

• some APIs may not support string to lat, long at the neighborhood level
• To solve this, the aggregate function could also be a "multisig", a location is valid if 3/4 API providers return a 200 response
• Another caveat is that some of these providers might be using the same underlying data provider
• This also is not mathematically deterministic, if APIs changed for some reason it would be hard to debug
• Perhaps there could be a running service that periodically geocoded and hash hashed a list of predetermined locations, if there was any deviation a flag could be raise

I'm not sure how FOAM works. The website is confusing, but three's another problem that I think FOAM might solve but this one deosn't: if you allowed people to enter locations at any granularity from Country to Neighborhood level is how long should the geohash be?

Fortunately many geocoding APIs have a "type" field. Which tells us how specific the location is.

We can map each provider's type to a defined geohash length

• "Berlin" -> [google_api] -> "city" -> [location_type_to_geohash_length_google] -> 7
• "Berlin" -> [openweather_api] -> "city" -> [location_type_to_geohash_length_openweather] -> 7
• "Berlin" -> [mapquest_api] -> "city" -> [location_type_to_geohash_length_openweather] -> 6

geohash_length_average = avg(7,7,6) -> 6.67 -> round(6.67) -> 6

The more I think about this the more I realize that there are workarounds for all the different edge cases. However, the fundamental challenge is that it's not mathetmatically deterministic. However, it could potentially be "deterministic enough".

Appendix

[gorli] fun fact (Görli is the name of both the Ethereum testnet and a park and (former?) train station in Berlin

[pricing_oracle] Chainlink pricing oracles