Vehicle Tracking with Fine-Grained GPS and Implications for Safety
Todd Humphreys, UT AustinD-STOP Symposium | March 2, 2015
Cell-Resolution Mobility Traces
Montjoye et al., “Unique in the Crowd: Privacy Bounds of Human Mobility” Nature, 2013
• 93% predictable (theoretical limit) (Song et al., 2010)
• highly unique: with only 4 random spatio-temporal points (1-hour time res.), 95% of traces can be identified (Montjoye et al., 2013)
At cell-level granularity, mobility traces are ...
Eagle and Pentland, “Eigenbehaviors: identifying structure in routine,” Behav Ecol Sociobiol, 2009
Behavior from Cell-Resolution Mobility Traces
Eigenbehavior analysis allows discovery of behaviors that are correlated across large gaps in time (e.g., I came in early so I can leave early and go to a show tonight).
Zang and Bolot, “Anonymization of location data does not work” MobiCom, 2011
The value of mobility traces goes astheir information content
From 25 million cell users across U.S. Sprint network ...
Cell-level: What are your daily patterns?
City-level: Where do you live?
Coarse GPS/WiFi-level:What is your approximate context?
Decimeter-level:What is your immediate present reality?
Why we may want our vehicles to know their position at sub-decimeter accuracy:
• safety and security for connected veh.• more efficient traffic flow • personal assistant; forecasting• proof of location