From: Eugen Leitl (eugen@leitl.org)
Date: Thu Oct 17 2002 - 04:32:42 MDT
On Wed, 16 Oct 2002, BillK wrote:
> My understanding of the present system is that the output is only blips
> on a screen. Bigger, faster blips would be vehicles, smaller blips -
We're not discussing the properties of a specific system, but reasonable
expectations of what is physically possible, and evaluation of trends.
I am not at all worried about current capabilities.
> pedestrians. (Remember the radar screens from the Gulf War showing
> hordes of dots streaming back to Iraq?)
> This is fine for isolated areas, say around power stations, defence
> establishments, etc. Any blips that appear could have a camera zoomed in
> on them. And mobile military want to use a similar system to spot enemy
> troops hiding in the countryside.
You can assume that a blip retains the identity as it follows a contiguous
trajectory (or semicontiguous, if you use kinetic models and blip
properties info to resolve ambiguities). If you have a complete recording
of all trajectories in the area, you can look at what happened after the
fact in the relevant area. As soon as this is a repeat offense, it will be
very hard to avoid making a pattern, even if you know how the system
works.
> If a Virginia shopping mall was being monitored when an incident
> occurred, then possibly a blip moving away at high speed would stand out
> as unusual. If it did then it could be tracked and followed from mast to
> mast until it stopped, or the tracking could be passed over to a police
> helicopter.
You're not tracking a single blip. You're tracking all the vehicles in the
area using distributed infrastructure. Once they're in the main memory, I
can track millions of individual objects in realtime on an off-shelf PC.
Do the math.
The hard part is gathering the information with an array of sensors. Lots
of smart sensors needs to be installed, serviced, and gathered info from.
Several trends (integration density, integrated software radios,
ultrawideband, cheap DSP, etc.) would seem to make blanket surveillance
affordable, especially if implemented as an incremental-cost rider on a
bona fide infrastructure.
> Similarly if a suspect was being monitored then a visual sighting of him
> leaving home could switch over to use the system to track his movements.
>
> But the overall recording of dots moving around a city is not a privacy
> problem at present.
If we're talking about more or less blanket coverage of large parts of the
city including all movements of vehicles and people this completely nukes
privacy.
> Now if they can add cell phone identification within two years, then
> recording the movements of every unique cell phone becomes possible.
You're describing capabilities of several years (a decade, if not more)
ago. There's a directional ID catcher gadget issued to individual LEOs.
The base stations know your position to within ~100 feet. Where do you
think mobile positional services obtain the info? It's not GPS yet.
> This data could indeed be stored for years. In theory, it would only
EU is pushing forwards to store all connection info for each person for a
year. This is different, but another data point showing the intent. Yes,
ve really are dat kurious.
> be examined by human eyes if a crime was being investigated.
Guess what, definition of a crime changes in time and in space. Look at
Saudia Arabia, look at kids at a rave.
> As Samantha said there would be far too much data for everyone's
> individual movements to be examined.
That's nonsense. 10^6 objects at a 1 Hz position refresh rate generate few
MBytes/s raw data flux. A current desktop is easily capable of processing
that stream in realtime. Same PC could easily hold TBytes of cheap RAID.
Stored naively, as a raw dump (of course, one wouldn't do that), that's a
FIFO almost two weeks deep. Then it's flushed out tiered archive storage.
Off-shelf clustering allows me to scale up that capacity by four orders of
magnitude. This means I can track all objects on Earth surface today with
a completely inconspicuous few 10 M$ facility. Allow for a decade or two
of technical advances, and continuation of current basic right erosion.
> But once the data existed, then it could be analysed by computer and
> lots of 'interesting' items might be highlighted. What would be
> considered 'interesting' would, of course, depend on which
> agency/department/team/bureaucrat was doing the analysis.
>
> And I think there is no chance that the public would be given access to
> the data files so that they could check on the movements of politicians
> / spouses / business competitors / etc.
You're getting it.
> Somehow I don't think I'll be carrying a cell phone in the future.
One could do traffic remixing in ad hoc meshes. Or make a habit of
swapping cellphones with friends (it's a good idea to do it with the
consumer cards you get at Albertsons and elswhere).
I'm not sure the phone identity gets changed when you swap out those
prepaid smart cards. Probably not.
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:17:38 MST