Towards Reliable Spatial Information in LBSNs

1. Towards Reliable Spatial Information in LBSNs Ke Zhang, Wei Jeng, Francis Fofie, KonstantinosPelechrinis, Prashant Krishnamurthy University of Pittsburgh ACM LBSN 2012 Pittsburgh, PA

2. Outline • Problem definition • Effects of fake check-ins • Fake check-in detection • Conclusion and future work

3. Location Sharing in LBSN

4. People can easily forge their whereabouts without proof of the locations… • Alter GPS’s API (FakeLocation) • Bypass localization module to manually check in a different venue than the actual one

5. Gain more virtual rewards Gain real rewards People usually use fake check-in to Mislead others

6. Our Goals and Contribution • Emphasize the effectsof fake spatial information in order to advocate the importance of identifyingfake location sharing • Provide a preliminary system based on location proof to detect fake check-ins

7. Fake Check-in Leads Monetary Losses… • Local businesses utilize LBSN as an inexpensive marketing channel for advertisement • Users can obtain special offers by checking-in to participating venues without their presence

8. Fake Check-in Results in Degraded Services.. • Noisy data will not guarantee high quality service • Foursquare provides recommendations by considering check-ins from • Users • Friends • Venues • Fake location information degrades the quality of service

9. Related Efforts • Foursquare provides the “cheater code” to minimize fake check-ins by imposing additional rules on users’ check-in frequency and speed • In our work we will utilize the primitives of location proofs

10. Our Scheme • We consider nearby fake check-ins: • Users check in to a locale that is nearby even if they are not physically present in it • Three assumptions: • The number of fake check-ins are less than the true ones • True check-ins are spatially within the venue; fake check-ins are largely distributed outside the venue • All devices have the same wireless capabilities

11. Location proofs User needs to provide location proof along every check-in • Received Signal Strength (RSS) vector measured from nearby WiFi APs Check-in points

12. Location Verification The LBSN provider utilizes recent k historical proofs provided by users who claims in the venue. • Apply density clustering to RSS vector space Noise Check-in points Clusters

13. Simulation Set Up • Venues are grouped into blocks of 6 and arranged in a 2D plane separated by streets • 90% of the venues are randomly assigned a WiFi AP • Users follow the RANK model to decide the next destination to check in • A user with a fake check-in will be positioned randomly outside the venue

14. Wireless Channel Model • Attenuation Factor Model for users to record RSS • : the signal strength at distance • : path loss exponent • : wall attenuation factor • : number of obstacles along the path • : noise with Gaussian

15. EvaluationResults • The performance is better when the wireless channel is stable • In a highly variable environment, our approach still performs efficiently • Detection works better with smaller number of fake check-ins

16. Conclusions • We bring the attention to the community on the effects of fake check-ins by analyzing various possible real-life situations • We design and evaluate via simulations a prototype detection system • Density clustering • Location proofs

17. Future Directions • Implement our system on real hardwareand examine • Real world performance • Effect of wireless hardware • Investigate different – more generic- approaches that do not depend on the assumptions made

18. Thank you