DECODE: Detecting Co-Moving Wireless Devices
Gayathri Chandrasekaran, Mesut Ali Ergin, Marco Gruteser, Rich Martin
(WINLAB, Rutgers University)
Jie Yang and Yingying Chen(Department of ECE, Stevens Institute of Technology)
Co-Movement
Moving Together in Close Proximity – (~ Few Meters)
Movement of – Multiple Transmitters held by the same person– Multiple Radios attached on the same device.
Motivation
GSM <SIM ID>Wifi <MAC Address> Bluetooth <Bluetooth ID>
How to map different Transmitters to a Device or Person ?
- Friend Finder Application
Motivation: Social Network MiningDynamic Human Social Network.
Viral Marketing(use pre-existing social networks to produce increases in brand awareness)
Localize and Estimate ?
(X1,Y1)
Localize and Estimate ?
Co- Mobile:Euclidean Distance <
Threshold
TRIVIAL ?
Localize and Estimate ?
Position Estimates Are Not Stable
Co-Mobile Transmitters
Approach
Detect Co-Movement by Observing RSSI– Co-Moving transmitters have Correlated fades– Dynamics of the environment are captured
Metric: Correlation Coefficient.
Properties of Co-Mobile Transmitters
Mobile: Small Scale Fading & Shadow FadingDifferent Small Scale FadingSimilar Shadow Fading
Non-
Co-Mobile Transmitters
Different Small Scale And Large Scale Fading
Correlation Co-Efficient
Pearson’s Product Moment Correlation Coefficient:-Determines the Linear Relationship between two Random Variables (Observed RSSI from 2 Devices)
Ranges between [-1, +1]+1 - Strong Positive Correlation0 - No Correlation-1 - Strong Negative Correlation
Experiment Goals
1. Quantify Co-Movement detection rates1. Within Same Technology2. Across Different Radio Technology
2. Quantify the Effect of Speed on the time taken to detect co-movement
Experimental Setup
5
Detecting Co-Movement (Similar Radio -
Wifi)
Impact of Speed on Co-Movement Detection Time
Threshold = 0.6
Conclusion: 50-100 Sec of Observation Required.
Impact of Speed on Co-Movement Detection Time
Other Observations
Sampling rates have a little effectWorks with mixed radio technologies (in 2.4GHz band) motes, wifiCorrelation Co-Efficient Threshold = 0.6 yields High Detection Rates and Low False PositivesDetect Mobile Periods by Observing Variance(>3) of RSSI over time.
Note: It only needs a single measurement node (base station)
Conclusion and Future WorkSimple and effective technique to Detecting Co-Mobile transmitters through signal strength analysis ( 2 minutes of Observation )Can work with just 1 receiver when the Localization infrastructure is Not in place
Future Work - Would a similar technique work outdoors using cellular signal strength information to detect co-moving cell phones?
QUESTIONS?
Thanks !
Determining Threshold (Τ)
Detecting Co-Movement (Different Radios)
Co-mobile Mote and Wifi
Mobility Detection using Variance.
Impact of Sampling Rate in Co-Movement Detection Time
DECODE System Design
Summary
Simple and effective technique to Detecting Co-Mobile transmitters through signal strength analysis ( 2 minutes of Observation )Can work with just 1 receiver when the Localization infrastructure is Not in placeCan Work Across radio TechnologiesLesser Time with increased Speed.Insensitive to Sampling Frequency.
Future Work
Would a similar technique work outdoors using cellular signal strength information to detect co-moving cell phones?
Application in Mind: Optimize phone Battery life by turning off the GPS for one of the Co-Mobile Devices.
How to Extend this to Detect Static Co-Located Transmitters?