1. Optimizing the Location Obfuscation in Location-Based Mobile Systems Iris Safaka Professor: Jean-Pierre Hubaux Tutor: Berker Agir Semester Project Security & Cooperation in Wireless Systems February 2012 1

2. Introduction Widespread use of location-based services (LBS) users reveal to a LBS provider personal location data Concern for location privacy protection unauthorized dissemination of location data inference of sensitive information about user Obfuscation: a common approach to protect privacy reveal to LBS provider coarse location information 2

3. Problem Statement Mobile user: reports set of locations between successive time instances Privacy concern in the presence of a passive eavesdropper Inspection of reported locations Knowledge on geographical area, user’s mobility model Processing of obfuscated traces -> tracking and localization attacks How does the level of privacy gets affected? 1234 5678 12 11109 13 141516 Location obfuscation 3

4. Observation: the way we choose the reported locations at each time instance influences the level of location privacy at the next time instance because: 1)Existence of spatial correlation between the reported locations at time t i and those at time t i+1 2)Transitions between locations are not equally probable An adversary could reduce her uncertainty about user’s real location using Bayesian inference Problem Statement 4

5. Losing Privacy - Example 1234 5678 Route: 2 -> 6 -> 7 56 78 5 12 1/2 timeline t0t0 t1t1 t2t2 1/4 1/2 10 1 1 0 0 0

6. Project Objective 6 Design a heuristic location obfuscation algorithm so as the deterioration of location privacy level between successive time instances is minimum consider observations (1) and (2) Intuition behind observations: The linkability graph should stay a full graph Select accessible and neighboring locations as fake ones The selected fake locations should be plausible Select locations that maximize

7. Which values to assign to the transitions between cells? History-based mobility model 7 Mobility Predictor Depends on the frequency of visiting next cell k starting from cell j during all past periods Direction-based mobility model

8. Combined mobility model 8 Mobility Predictor whereis the combination parameter We assign values to transitions using the following whereis the non-moving probability

9. The heuristic algorithm - Example 1234 5678 Route: 1->2->3 12 {1,2,5,6} timeline t0t0 t1t1 t2t2 {1,2,5,6} {1,2,3,5,6,7} {1,2,5,6} 26 {1,2,3,5,6,7} 36 9 1/2 Pr(6, t 1 ) Pr(2, t 1 ) Pr(6, t 2 )Pr(3, t 2 )

10. Figure 1: Accessibility map and trace generators 10 Evaluation Framework Figure 2: An example of a trace

11. Figure 3: Location obfuscation and location privacy measurement 11 Evaluation Framework

12. Setup parameters 20 mobile users 10 consecutive time instances Location obfuscation parameter c=2 and c=3 12 Experimental Results Figure 4: Location privacy level for different mobility predictors

13. 13 Experimental Results Figure 5: Comparison of effectiveness between heuristic and random algorithm (a) For c=2(b) For c=3

14. Contribution A heuristic algorithm for location obfuscation that aims to minimize deterioration of privacy level between successive time instances in a mobile setup Implementation of an evaluation framework and experimentation Future Work Identify and formally prove necessary and sufficient conditions Extend heuristic for non-constant user speed and location obfuscation parameter Experimentation on real-world traces 14 Conclusion