Nowhere to hide? Mix-Zones for Private Pseudonym Change using Chaff - - PowerPoint PPT Presentation

Nowhere to hide? Mix-Zones for Private Pseudonym Change using Chaff Vehicles

Christian Vaas1, Mohammad Khodaei2, Panos Papadimitratos2, Ivan Martinovic1

University of Oxford1, KTH, Stockholm2

VNC, 6 December 2018

Motivation: Vehicle Tracking

2

VNC 2018

• Static credentials to sign CAM

messages

• Wireless eavesdropper
• Attacker records full route

➢ Identify individuals [1] ➢ Movement profiles

Pseudonym Change

3

VNC 2018

• Frequently changing credentials
• Attacker records route segments

➢ How to make segments unlinkable?

Segment Unlinkability

4

VNC 2018

1. Syntactic linking ➢ Synchronization 2. Semantic linking ➢ Obfuscate CAMs

t0 t1

Mix-zones: Principle

• Covers intersections
• Prevent pseudonym linking
• Obfuscate CAMs
• Silent or encrypted periods
• Private change of ECDSA

credentials

➢ Recording entry-exit pairs

5

VNC 2018 Q Y P B C A

?

R

Crypto Mix-zones: Problem

6

VNC 2018

• Dependence on vehicle

availability

• Low traffic regions
• Low traffic hours
• Driver population
• Arrival timings

➢ Correlation attack

Q A

Our Solution

7

VNC 2018

➢ Chaff vehicles

• Substitute for real vehicles
• RSUs generate chaff CAMs
• CAMs signed with chaff

pseudonyms

• CAMs broadcast by RSUs

and OBUs

• Must not impair safety

Q Y P

Y’ P’

A

?

R

R’

D B A C

Real vehicle v Real vehicle v' Chaff vehicles

D B A C

Real vehicle v Real vehicle v' Chaff vehicles Real vehicle v Real vehicle v'

A

Chaff-based CMIX Scheme

8

VNC 2018

Protocols & Services 1. Key provisioning 2. Chaff-trace generation 3. Chaff notification

• D to B

4. Filter update ➢ Safety preserving ➢ Maximize mixing

Real vehicle v Real vehicle v' Chaff vehicles

A

Simulation Environment

9

VNC 2018

LuST

Luxembourg SUMO Traffic Scenario

PREXT

Privacy Extension for Veins VANET Simulator

CMIX

Mix-Zones for Location Privacy in Vehicular Networks

Simulation Scenarios

10

VNC 2018

• Three different areas
• Suburban – low traffic
• Residential – medium traffic
• Central – high traffic
• Encryption radius fixed
• Tracking probability based metric

𝑄 𝑈

𝑤 = σ𝑗=0 𝑛 𝑞𝑤 𝑜𝑗 ∗ 𝑚 𝑓𝑗

|𝑈

𝑤|

A Q

𝑜𝑗 𝑓𝑗

• Encryption radius 50 m
• Attacker strength 100%

Simulation Results

11

VNC 2018

0:00 7:30 13:00 18:30 24:00 Simulation tim e [HH:mm] 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Normalized pseudonym change exposure Central Suburban Residential CMIX scheme 0:00 7:30 13:00 18:30 24:00 Simulation time [HH:mm] 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Normalized pseudonym change exposure 1000 2000 3000 4000 Active real vehicles [# ] Central Suburban Traffic density Residential CMIX scheme 0:00 7:30 13:00 18:30 24:00 Simulation time [HH:mm] 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Normalized pseudonym change exposure 1000 2000 3000 4000 Active real vehicles [# ] Central Suburban Traffic density Residential Chaff-based scheme CMIX scheme

Simulation Results

12

VNC 2018

• Encryption radius 50-250 m
• Attacker strength 30%/100%

Mix-zone encryption rad ius [m]

Normalized pseudonym change exposure 0.0 0.2 0.4 0.6 0.8 CMIX schem e Chaff-based scheme 50 100 150 200 250

Residential - 30%

50 100 150 200 250 0.0 0.2 0.4 0.6 0.8 CMIX scheme Chaff-based scheme

Residential - 100%

System Feasibility

13

VNC 2018

Mix-zone Encryption Radius [m] 50 100 150 200 250

• Max. Active Chaff Pseudonyms [#]

68 176 165 99 66

• Max. CAM Generation [msg/s]

240 848 1321 831 634

• 30 * 60 * 176 = 316,800 chaff pseudonyms
• Cuckoo Filter with 3.63 MB

✓ Transmission speed 6 Mbit/s in IEEE 802.11p

• Generate 6742 ECDSA signatures per second

✓ NEXCOM (Dual-core 1.66 GHz, 1GB memory) with crypto module

D B

Real vehicle v Real vehicle v' Chaff vehicles

14

VNC 2018

• New pseudonym change strategy based on chaff vehicles and

chaff messages

• Independent of operation area, mix-zone encryption radius, time
• f day, and driver population
• System performance: up to 76% improvement over CMIX
• Preserves safety application functionality
• Acceptable resource requirements
• Resilience against internal attackers
• Impact of honest-but-curious VPKI entities

Conclusion Future Work

Thank you!

christian.vaas@cs.ox.ac.uk

Nowhere to hide? Mix-Zones for Private Pseudonym Change using Chaff Vehicles

References

16

VNC 2018

[1] Golle, P., & Partridge, K. (2009, May). On the anonymity of home/work location pairs. In International Conference on Pervasive Computing (pp. 390-397). Springer, Berlin, Heidelberg. [2] L. Codeca, R. Frank, S. Faye and T. Engel, "Luxembourg SUMO Traffic (LuST) Scenario: Traffic Demand Evaluation" in IEEE Intelligent Transportation Systems Magazine, vol. 9, no. 2,

• pp. 52-63, Summer 2017.

[3] PREXT: Privacy Extension for Veins VANET Simulator", IEEE Vehicular Networking Conference (VNC), Dec. 2016, Columbus, Ohio, USA [4] Freudiger, J., Raya, M., Félegyházi, M., Papadimitratos, P., & Hubaux, J. P. (2007). Mix- zones for location privacy in vehicular networks. In ACM Workshop on Wireless Networking for Intelligent Transportation Systems (WiN-ITS) (No. LCA-CONF-2007-016).

Simulation Results

17

VNC 2018

0:00 4:00 8:00 12:00 16:00 20:00 24:00 Simulation time [HH:mm] 0.00 0.01 0.02 0.03 0.04 0.05 0.06 Chaff vehicles per real vehicle 1000 2000 3000 4000 Active real vehicles [# ] Traffic density Central Suburban Residential

Simulation Parameters

18

VNC 2018 Area 1.31 km2 0.61 km2 1.38 km2 Junctions 69 34 61 Mix-zones 31 18 28

• Avg. Number of

vehicles per zone 1825 4631 6500