Elastic Pathing: Your Speed Is Enough to Track You Presented by Ali
Motivation Insurance companies claim to collect only speed data Attracting users to install monitoring device Example: Progressive: Snapshot device
Problem It is possible to track you by knowing starting point and driving speed with timestamps Insurance companies know your starting point, home address Speed data not considered confidential by insurance companies
Challenge Possibilities of multiple paths Hard to determine turning direction
Elastic Pathing Algorithm: Overview OpenStreetMap
Elastic Pathing Algorithm: Error Detection Errors happen when: stopping in the midway speed is too fast to make turns Path correction: Expanding (stretching) Compressing Degree of correction affects the path score
Elastic Pathing: Assumptions Drivers will stop only at traffic lights and stop signs Each vehicle has physical limitation – turning radius No vehicle can make turns at high speed
Elastic Pathing Algorithm Chooses the path with smallest error Checks for max. and min speeds Makes a “landmark” when speed trace and road data match OR: A vehicle has come to a stop at an intersection Sorts all possible paths
Elastic Pathing Algorithm: Example
Elastic Pathing Algorithm: Accuracy New Jersey dataset 14% traces with error less than 0.16 miles – 250 m 24% traces with error less than 0.31 – 500 m Seattle dataset 13% traces with error less than 0.16 miles – 250 m 26% traces with error less than 0.31 miles – 500 m
Real-world Sample
Conclusion Accuracy does not go down with trip distance The algorithm depends on the driving habits Distinguishing between two different roads having similar features
Demo