1 preserving privacy in gps traces via uncertainty-aware path cloaking by: baik hoh, marco gruteser,...
TRANSCRIPT
![Page 1: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/1.jpg)
1
Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking
by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf AlrabadyACM CCS '07
Presentation: Martin AzizyanECE 256, Spring 09
Duke University
![Page 2: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/2.jpg)
2
Overview
Introduction Problem Previous work Proposed methods Evaluation Discussion
![Page 3: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/3.jpg)
3
Introduction
Emerging use for aggregate location traces Automotive traffic monitoring City planning
Privacy a big issue Individuals can be “followed” with their traces
Existing techniques have drawbacks Either sacrifice data accuracy, or anonymity
![Page 4: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/4.jpg)
4
Traffic monitoring
Goal: estimate travel time for routes “Probe vehicles” report real-time position and speed Data stored in central database for analysis
Both real-time and historical
![Page 5: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/5.jpg)
5
Traffic monitoring
Requires high spacial accuracy Parallel roads may be only 10m apart Thus, individuals can be tracked with high accuracy
In area of high density traffic, not an issue Can't track one person in a crowd
Privacy must also be guaranteed in low density Though data from low-traffic routes not as important
![Page 6: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/6.jpg)
6
Existing privacy algorithms (1)
K-anonymity Guarantees degree of anonymity Very low accuracy
![Page 7: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/7.jpg)
7
Existing privacy algorithms (2)
Best effort Exploit confusion from multiple crossing paths
![Page 8: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/8.jpg)
8
Existing privacy algorithms (2)
Best effort Tang et al.
Subsampling
![Page 9: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/9.jpg)
9
Existing privacy algorithms (2)
Best effort Tang et al.
Subsampling
![Page 10: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/10.jpg)
10
Existing privacy algorithms (2)
Best effort Tang et al.
Subsampling Non-uniform subsampling also explored
Suppress information in high-density areas Unclear worst-case privacy guarantees
Individual users still at risk
![Page 11: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/11.jpg)
11
Trace privacy metric
Given trace, determine degree of privacy Mean Time To Confusion (MTTC)
Time adversary can correctly follow a trace Need Adversary model
Last position + heading ~ current position Calculate Tracking Uncertainty H due to confusion If H > a threshold, then assume trace lost
MTTC depends on threshold for H
![Page 12: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/12.jpg)
12
Proposed algorithm
Parameter: maximum time to confusion Longest time interval a trace can be followed Also need to set maximum uncertainty level
Divide into time slots For each sample in a time slot, check:
Time since last point of confusion < max Tracking uncertainty > min If either satisfied, release sample (make available)
![Page 13: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/13.jpg)
13
Possible modifications
Algorithm not specific to one adversary model Independent tracking uncertainty calculation
Reacquisition tracking model Adversary can skip over some points of confusion Minor modifications to algorithm necessary
![Page 14: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/14.jpg)
14
Experimental setup
Data Collected GPS traces from 233 vehicles
Sample includes timestamp, coordinates, velocity and heading
Experiments performed on 24 hour traces With 500 and 2000 probe vehicles One vehicle's traces from 24 hour periods simulate
multiple vehicles
![Page 15: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/15.jpg)
15
Experimental setup
Evaluation metrics Maximum and median time to confusion (TTC) Relative weighted road coverage
Each sample assigned weight based on number of samples in its area
Quality of sample set = sum of sample weights
![Page 16: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/16.jpg)
16
Results
High-density scenario (2000 vehicles) Without reacquisition
![Page 17: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/17.jpg)
17
Results
High-density scenario (2000 vehicles) With reacquisition
![Page 18: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/18.jpg)
18
Results
Low density scenario (500 vehicles)
Without reacquisition With reacquisition
![Page 19: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/19.jpg)
19
QoS analysis
Samples kept: uncertainty-aware algorithm v.s. random sampling
![Page 20: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/20.jpg)
20
QoS analysis
Relative weighted road coverage No significant change after executing algorithm
![Page 21: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/21.jpg)
21
QoS analysis
Maximum TTC vs. weighted road coverage
Without reacquisition With reacquisition
![Page 22: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/22.jpg)
22
Discussion
Map-based tracking Roads not a continuous 2D space Adversary can assign probabilities more intelligently
A priori knowledge Tracking select individual easier than data mining
Trust in central location server Fully distributed approach seems infeasible Hybrid approach more likely Inform vehicle of probe density in their area
![Page 23: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/23.jpg)
23
The End
![Page 24: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/24.jpg)
24
5.1 snapshots:
![Page 25: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/25.jpg)
25
Proposed algorithm
Processes with time slots Reveals sample if confusion
![Page 26: 1 Preserving Privacy in GPS Traces via Uncertainty-Aware Path Cloaking by: Baik Hoh, Marco Gruteser, Hui Xiong, Ansaf Alrabady ACM CCS '07 Presentation:](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649dc45503460f94ab6d37/html5/thumbnails/26.jpg)
26
Existing privacy algorithms (2)
Best effort Exploit confusion from multiple crossing paths