- From: Doug Turner <doug.turner@gmail.com>
- Date: Tue, 10 Jun 2008 22:18:57 -0700
- To: timeless@gmail.com
- Cc: public-geolocation@w3.org
On Jun 10, 2008, at 9:52 PM, timeless wrote: > On Jun 10, 2008, at 7:43 PM, Kartikaya Gupta wrote: >> It seems to me that any approach of this sort would be vulnerable to >> attacks. As somebody mentioned in a previous post, random fuzzing >> can be >> defeated by doing multiple requests and averaging the results. > >> In the snap-to-grid approach I think you're describing, a more >> precise >> position can be pinpointed if you poll the location repeatedly and >> record >> the exact moment you switch from one grid-line to another. i.e. If >> your >> fuzzing reduces precision by rounding down, say from 3.19 to 3.1, >> then the >> moment that value switches to 3.2 means the actual location has >> switched >> from 3.19 to 3.20, and you have your precision back. > > On Wed, Jun 11, 2008 at 7:09 AM, Doug Turner <doug.turner@gmail.com> > wrote: >> Why can't you just remove/round precision from the lat/long? >> >> For example, this: 37.41857,-122.08769, becomes 37.4, -122.1 > > this was already explained, in fact in the precise message to which > you replied. > we're not talking about users sending one fixed data point to a > server once. > > we're talking about users engaged in a dialog where their client will > send updates, and as soon as the update needs to cross from one > truncated value to another, the service can determine with precision > where the user is (one axis at a time, true, but still with > precision). yeah, I guess I don't see that happening. I think we are talking about rounding of a bunch of places on the lat-long. for example, everything in 37.4, -122.1 is a large area. When I move to the next box, the system we be able to tell that I am at the board between the two regions. Is that the worry? Doug
Received on Wednesday, 11 June 2008 05:41:02 UTC