Wireless Location Privacy: Radiometric Breaches and Defenses Marco - PowerPoint PPT Presentation

Wireless Location Privacy: Radiometric Breaches and Defenses Marco Gruteser WINLAB

Trends Always-on, transmitting Low-cost software (controlled) radios background apps (e.g., GNU radio)

Removing Identifiers at or above Bit-Level: Example Anonymization Access Control GPS Satellite Location Cellular Traffic Estimation Proxy Data mining and Service logging Provider Vehicle ID | timestamp | Lon | Lat | Speed | Heading ------------------------------------------------------------------ 254,18-oct-2006 10:11:12,-85.3452,42.4928,42.18,135 372,18-oct-2006 10:11:12,-85.3427,42.4898,63.72,100 182,18-oct-2006 10:11:12,-85.4092,42.4726,50.15,75 Probe 254,18-oct-2006 10:12:12,-85.3462,42.4998,45.18,135 Vehicles 372,18-oct-2006 10:12:12,-85.3512,42.4944,60.01,185 182,18-oct-2006 10:12:12,-85.4102,42.4753,45.88,235 … 254,18-oct-2006 10:21:12,-85.3856,42.5129,45.67,135 Anonymous Trace log files

Privacy Im plications at the Signal Level? Beyond Localization … • Possible Breaches – Identify transmitters? – Infer information other than location? • Protections – Prevent long-term tracking? – Thwart localization?

Recent Controversy about Human Mobility Article Demonstrates Location Privacy Sensitivities • Analyzed human mobility patterns from cell phone hand off data – 100,000 users over 6 months • Coarse data – Cell tower location recorded for each call / message – Average towers covered 3 km^ 2 • Example of typical assumptions about location inference from signals

Identifying Transmitters via Radiometric Signatures with Suman Banerjee (MobiCom’08)

Waveform Impairments in Analog Frontend

Transmitter Identification via Classification • Training phase: collect fingerprint (waveform error metrics) of each transmitter • Identification phase: measure error metrics for candidate transmitter and use classification algorithm to match with training set

ORBIT

Identification Results • Using ORBIT testbed radios and vector signal analyzer for data collection • K-Nearest Neighbor and Support Vector Machine classifiers

Identifying Co-location (While Moving) without Location With Rich Martin, Yingying Chen (MASS 08)

Motivating Applications • Co-location can reveal social relationships among device owners

Properties of Co-Mobile Transmitters Small scale fading � Fast fading differs but slow fading similar

Non- Co-Mobile Transmitters Both large and small scale fading differs

Detection via Filtering and Time- Series Correlation � Pearson’s Product Moment Correlation Coefficient:- Determines the Linear Relationship between two Random Variables (Observed RSSI from 2 Devices) � Ranges between [-1, +1] – 0 : No Correlation – +1 : Strong Positive Correlation – -1 : Strong Negative Correlation

Experimental Setup

Impact of Speed on Co-Movement Detection Time Correlation-Coefficient Threshold >= 0.6 Conclusion: 50-100 Sec of Observation Required.

Thwarting Tracking: Applying Path Cloaking with Hui Xiong (CCS 07)

Inference/ Insider Attacks Compromise Location . Privacy . . . . . Still insider breaches and remote break-ins possible . . . . . . Re-identification . of traces . . through data . analysis Tracking algorithms recover individual trace [Hoh05] (Median trip time only 15min) Anonymous Home Trace log Identification files [Hoh06] Location may be precise enough to identify home

Stronger De-identification for Location Traces: Filtering based on Tracking Model t=T 2 t=T 1 d 1 t=T 2 d 2 t=T 1 d 1 t=T 2 d 3 d 2 t=T 2 d 3 t=T 2 Low Entropy High Entropy t=T 2 Low Uncertainty High Uncertainty Uncertainty Normalization

Thwarting Precise Localization Preliminary Work

(X1,Y1) ‏ 3db Power Control?

3db Power Control Provides Little Benefit (X1,Y1) ‏ 3db 3db 3db

Cooperation can yield Asymmetric 3db Signal Changes 1db 0db

Proof-of-Concept Result 18 18 Victim Node 17 12dBm 17 16 Cooperator−1 16 15 Decoy Location −1 15 14 14 13 13 12 12 6dBm Grid − y Grid − y 11 11 10 10 6dBm 9 9 8 8 7 7 1dBm 6 6 Victim Node 5 1dBm 5 Cooperator−2 4 4 3 12dBm 3 Decoy Location−2 2 2 4 5 6 7 8 9 10 11 12 13 14 15 16 4 5 6 7 8 9 10 11 12 13 14 15 16 Grid − x Grid − x

t=T 2 t=T 2 t=T 2 d 1 d 2 d 3 t=T 1 3db Summary 1db 0db

Thank you