1 of 30

Factor Graph-based Spoofing Mitigation using the Chimera Signal Enhancement

Ashwin V. Kanhere, Tara Mina, Akshay Shetty and Grace Gao

ION GNSS+ 2022

2 of 30

Civilian GNSS Signals are Vulnerable to Spoofing Attacks

  • Civilian GPS signals used in many critical infrastructure areas[1] but vulnerable to attack:
      • Unencrypted
      • Structure publicly known

​

  • Approximately 10,000 spoofing events documented in Russian territories from 2016-2019[2]

​

  • “Circle spoofing” attacks documented in Norway, Malaysia and Nigerian coast[3]

1

[2] Above Us Only Stars, C4ADS 2019

[1] Report on PNT capabilities of GPS, US DHS 2017

Civilian GPS signals are used in critical infrastructure such as communication, finance, emergency services and the power grid

[3] Goward, GPS World 2020

Documented instances of “circle spoofing” across the world, spoofed to Western North American coast[3]

3 of 30

Chimera Signal Enhancement would�Provide Security to Civilian Signals[4]

  • Spoofer cannot perfectly predict authentication markers before broadcast

​

  • Decrypt received markers using key to authenticate GPS measurements

​

  • Authenticate GPS measurements with compatible receiver

​

  • High latency between authentications

2

[4] Anderson et al., ION GNSS+ 2017

Slow channel

Fast channel

Markers

Markers

Key from data signal

Authentication every �3 mins

Key from external signal

Authentication every �1.5 or 6 secs

4 of 30

Self-Contained Odometry can Augment �Low Latency GPS Measurements

  •  

3

True trajectory

Odometry-only trajectory

Odometry + GPS trajectory

Spoofed GPS trajectory

5 of 30

Self-Contained Odometry Sensors can�Mitigate Attacks Between Authentication

  • When used with Chimera-enhanced GPS measurements:
      • Coast between Chimera authentications [5]
      • Detect spoofing between Chimera authentications[6]

​

  • Combining multiple sensors is a filtering or batch optimization problem

​

4

True trajectory

Odometry-only trajectory

Odometry + authentic GPS �trajectory

Spoofed GPS trajectory

Unauthenticated GPS

Authenticated GPS

[5] Esswein et al., ION GNSS+ 2021

[6] Mina et al., ION GNSS+ 2021

Spoofed GPS and odometry deviate

This work: Combine self-contained odometry with Chimera-enhanced GPS measurements to mitigate effect of spoofing attacks between authentication

6 of 30

Filtering and Optimization Approaches�to Mitigate Spoofing

Session F6: GNSS Receivers Robust to Vulnerabilities

​

GPS Spoofing-Resilient Filtering with Chimera and Self-Contained Odometry

Tara Mina, Ashwin V. Kanhere, Akshay Shetty and Grace Gao

We explored both filtering and batch optimization approaches this summer:

Session B6b: Spectrum: Protection and Optimization

​

Factor-Graph-based Spoofing Mitigation using Chimera Signal Enhancement

Ashwin V. Kanhere, Tara Mina, Akshay Shetty and Grace Gao

​

5

THIS PRESENTATION

7 of 30

Contributions

​

  1. We designed a factor graph optimization-based approach to mitigate spoofing�
      • Formulate a robust cost function that includes intuition of spoofing attacks�
      • Incorporate Chimera signal authentication as a loop closure to include authentication for past state estimates�

​

  • We experimentally validated, via simulation, spoofing mitigation performance and improvement in state estimates by including Chimera authentication

6

8 of 30

Factor Graph-based Spoofing Mitigation �using the Chimera Signal Enhancement

  • Background: Factor Graph Optimization
      • Factor Graphs
      • Robust Factor Graph Optimization�
  • Proposed FGO-based Spoofing Mitigation with Chimera
      • Graph states and Measurements
      • Chimera authenticated Loop Closure
      • Robust Loss Formulation�
  • Experimental Validation�
  • Summary

7

9 of 30

Factor Graphs Perform Batch�Optimization For State Estimation

  •  

8

10 of 30

Factor Graphs Perform Batch�Optimization For State Estimation

  •  

9

 

Iteration:

0

1

2

11 of 30

Factor Graphs Perform Batch�Optimization For State Estimation

  •  

10

New measurements received

Old states and measurements removed

12 of 30

Factor Graph Optimization (FGO) �Loss Functions

  •  

11

[8] Sunderhauf, PhD Thesis(2012)

[7] Wen et al. NAVIGATION (2021)

Add over all graph edges in window

Received measurement

Expected measurement

Measurement covariance

Error vector

Switchable constraint value

Regularization function

13 of 30

Factor Graph With Switchable �Constraint

  •  

12

14 of 30

Factor Graph-based Spoofing Mitigation �using the Chimera Signal Enhancement

  • Background: Factor Graph Optimization
      • Factor Graphs
      • Robust Factor Graph Optimization�
  • Proposed FGO-based Spoofing Mitigation with Chimera
      • Graph States and Measurements
      • Chimera Authenticated Loop Closure
      • Robust Loss Formulation�
  • Experimental Validation�
  • Summary

13

15 of 30

FGO-based Spoofing Mitigation with �Chimera: Formulation

  •  

14

16 of 30

FGO-based Spoofing Mitigation with �Chimera: States

  •  

15

17 of 30

FGO-based Spoofing Mitigation with �Chimera: Measurements

  •  

16

[9] Lupton et al. IEEE TRO, 2012

18 of 30

FGO-based Spoofing Mitigation with�Chimera: Authentication Loop Closure

  •  

17

19 of 30

FGO-based Spoofing Mitigation with�Chimera: Loss Function

Naïve loss formulation

18

IMU odometry loss term

Wheel encoder odometry loss term

GPS position loss term

 

 

 

 

20 of 30

FGO-based Spoofing Mitigation with�Chimera: Loss Function

Spoofing robust loss formulation

19

IMU odometry loss term

Wheel encoder odometry loss term

Weighted GPS position loss term

 

 

 

 

 

Switching constraint regularization

 

Penalize multiple switches

21 of 30

Factor Graph-based Spoofing Mitigation �using the Chimera Signal Enhancement

  • Background: Factor Graph Optimization
      • Factor Graphs
      • Robust Factor Graph Optimization�
  • Proposed FGO-based Spoofing Mitigation with Chimera
      • Graph States and Measurements
      • Chimera Authenticated Loop Closure
      • Robust Loss Formulation�
  • Experimental Validation�
  • Summary

20

22 of 30

Simulation Setup And Parameters

  •  

21

Sensor

Update Rate

​

GPS

​

​

IMU

​

​

Wheel Encoder

​

​

[9] Lupton et al. IEEE TRO, 2012

23 of 30

Simulation Setup And Parameters

  •  

22

24 of 30

Accuracy Comparison: �Robust FGO vs. Naïve FGO

23

Method/Trajectory

Mean Error�(in meters)

Max error�(in meters)

Robust FGO/Nominal

​

​

Robust FGO/Spoofed

​

​

Naïve FGO/Nominal

​

​

Naïve FGO/Spoofed

​

​

Spoofing start

Robust FGO mitigates spoofing, improving accuracy over naïve FGO

25 of 30

Switching Constraint Value Under�Nominal And Spoofed Conditions

24

 

Authentication propagates to previous times in window on authentication

 

Lack of Chimera authentication during spoofing prevents loop closure

 

Spoofing start

26 of 30

Performance with and without Chimera Authentication for Nominal Trajectory

25

Method/Trajectory

Mean Error�(in meters)

Max error�(in meters)

Robust FGO/Nominal�With Chimera

​

​

Robust FGO/Nominal

Without Chimera

​

​

Using Chimera authentication results in accuracy improvement

27 of 30

Factor Graph-based Spoofing Mitigation �using the Chimera Signal Enhancement

  • Background: Factor Graph Optimization
      • Factor Graphs
      • Robust Factor Graph Optimization�
  • Proposed FGO-based Spoofing Mitigation with Chimera
      • Graph States and Measurements
      • Chimera Authenticated Loop Closure
      • Robust Loss Formulation�
  • Experimental Validation�
  • Summary

26

28 of 30

Summary

  • We designed a robust FGO-based approach using Chimera-enhanced GPS signals to mitigate spoofing attacks

​

  • Our proposed method leverages properties of spoofing attacks in a spoofing aware switching cost constraint

​

  • We incorporate Chimera signal authentication as a loop closure to back-propagate effect of authentication for improved state estimates

​

  • Our proposed architecture was validated on simulated data to demonstrate spoofing mitigation and effectiveness of loop closure

27

29 of 30

Acknowledgments

  • Sriramya Bhamidipati, Derek Knowles, Shubh Gupta and Joseph Lucero for insightful discussions and feedback on this presentation

​

  • Other Stanford NAV Lab members

​

  • This material is based upon work supported by the Air Force Research Lab (AFRL) under grant number FA9453-20-1-0002

28

30 of 30

Summary

  • We designed a robust FGO-based approach using Chimera-enhanced GPS signals to mitigate spoofing attacks

​

  • Our proposed method leverages properties of spoofing attacks in a spoofing aware switching cost constraint

​

  • We incorporate Chimera signal authentication as a loop closure to back-propagate effect of authentication for improved state estimates

​

  • Our proposed architecture was validated on simulated data to demonstrate spoofing mitigation and effectiveness of loop closure

29