Announcements Today: Last lecture , special topic on smart - - PowerPoint PPT Presentation

▶
announcements
SMART_READER_LITE
LIVE PREVIEW

Announcements Today: Last lecture , special topic on smart - - PowerPoint PPT Presentation

Sep 23, 2023 227 likes •771 views

Announcements Today: Last lecture , special topic on smart transportation security Attention: Its within the scope of final exam Final exam: 12/12, 1:30-3:30 PM Should be in this class room (HSLH 100A) Bring your photo ID with

slide-1
SLIDE 1

1

Today: Last lecture, special topic on smart transportation security

  • Attention: It’s within the scope of final exam

Final exam: 12/12, 1:30-3:30 PM

  • Should be in this class room (HSLH 100A)
  • Bring your photo ID with you

Announcements

slide-2
SLIDE 2

DNS: Domain Name Service

Client Local DNS recursive resolver root & edu DNS server uci.edu DNS server

www.ics.uci.edu

ics.uci.edu DNS server

DNS maps symbolic names to numeric IP addresses

(for example, www.uci.edu ↔ 128.195.188.233)

2

slide-3
SLIDE 3

Cached Lookup Example

Client Local DNS recursive resolver root & edu DNS server uci.edu DNS server ics.uci.edu DNS server

ftp.ics.uci.edu

3

slide-4
SLIDE 4

DNS “Authentication”

Client Local DNS recursive resolver root & edu DNS server uci.edu DNS server

www.ics.uci.edu

ics.uci.edu DNS server Request contains random 16-bit transaction id  TXID Response accepted if TXID is the same Stays in cache for a long time (TTL)

4

slide-5
SLIDE 5

DNS Spoofing / DNS Cache Poisoning

Client Local resolver ns.foo.com DNS server

www.foo.com

Trick client into looking up www.foo.com (how?) Guess TXID, www.foo.com is at 6.6.6.6

6.6.6.6

Another guess, www.foo.com is at 6.6.6.6 Another guess, www.foo.com is at 6.6.6.6

Several opportunities to win the race If attacker loses, has to wait until TTL expires … but can try again with host1.foo.com, host2.foo.com, etc. … but what’s the point of hijacking host2.foo.com?

5

slide-6
SLIDE 6

DNS Spoofing / DNS Cache Poisoning

Client Local resolver ns.foo.com DNS server

<random>.foo.com

Trick client into looking up <random>.foo.com Guessed TXID, very long TTL I don’t know where <random>.foo.com is Ask the authoritative server at www.foo.com It lives at 6.6.6.6

6.6.6.6 If attacker wins, future DNS requests for www.foo.com will go to 6.6.6.6 The cache is now poisoned… for a very long time! No need to win future races!

[Kaminsky]

www.foo.com

6

slide-7
SLIDE 7

DNSSEC

  • Goals: authentication and integrity of DNS

requests and responses

  • PK-DNSSEC (public key)

– DNS server signs its data (can be done in advance) – How do other servers learn the public key?

MORE INFO: http://www.dnssec.net/presentations

7

slide-8
SLIDE 8

Lecture 17 CS 134 Smart Transportation Security

Qi Alfred Chen Department of Computer Science

slide-9
SLIDE 9

Recent interest: Autonomy software security in smart transportation

9

Connected Vehicle (CV) Autonomous Vehicle (AV)

slide-10
SLIDE 10

Recent interest: Autonomy software security in smart transportation

10

Connected Vehicle (CV) Autonomous Vehicle (AV)

slide-11
SLIDE 11

Recent interest: Autonomy software security in smart transportation

11

Connected Vehicle (CV) Autonomous Vehicle (AV)

slide-12
SLIDE 12

Recent interest: Autonomy software security in smart transportation

12

Connected Vehicle (CV) Autonomous Vehicle (AV)

slide-13
SLIDE 13

Recent interest: Autonomy software security in smart transportation

13

Connected Vehicle (CV) Autonomous Vehicle (AV)

Autonomy software

slide-14
SLIDE 14

Recent interest: Autonomy software security in smart transportation

14

Connected Vehicle (CV) Autonomous Vehicle (AV)

[ISOC NDSS’18] First software security analysis of a CV-based transportation system [ACM CCS’19] First software security analysis of LiDAR-based AV perception

Autonomy software

slide-15
SLIDE 15

Recent interest: Autonomy software security in smart transportation

15

Connected Vehicle (CV) Autonomous Vehicle (AV)

[ISOC NDSS’18] First software security analysis of a CV-based transportation system [ACM CCS’19] First software security analysis of LiDAR-based AV perception

slide-16
SLIDE 16

CV = Connected Vehicle OBU = On-Board Unit RSU = Road-Side Unit

Background: Connected Vehicle technology

  • Wirelessly connect vehicles & infrastructure to

dramatically improve mobility & safety

  • Will soon transform transportation systems today

– 2016.9, USDOT launched CV Pilot Program

16

RSU OBU

CV technology

Under deployment

slide-17
SLIDE 17

First security analysis of CV-based transp.

  • Target: Intelligent Traffic Signal System (I-SIG)

– Use real-time CV data for intelligent signal control – USDOT sponsored design & impl. – Fully implemented & tested in Anthem, AZ, & Palo Alto, CA

  • ~30% reduction in total vehicle delay

– Under deployment in NYC and Tampa, FL

17

I-SIG

Control Real-time CV data

RSU

CV = Connected Vehicle OBU = On-Board Unit RSU = Road-Side Unit

slide-18
SLIDE 18

Threat model

  • Malicious vehicle owners deliberately control the

OBU to send spoofed data

– OBU is compromised physically1, wirelessly2, or by malware3

  • Can only spoof data, e.g., location & speed

– Can’t spoof identity due to USDOT’s vehicle certificate system

18

I-SIG

Influence signal control Spoofed CV data

RSU Malicious vehicle owner

Control Real-time CV data

2 Checkoway et al.@Usenix Security'11 1 Koscher et al.@IEEE S&P’10 3 Mazloom et al.@UsenixWOOT’16

slide-19
SLIDE 19

Attack goals

19

Traffic congestion

Increase total delay of vehicles in the intersection

Personal gain

Minimize attacker’s travel time (at the cost of others’)

slide-20
SLIDE 20

Attack goals

20

Traffic congestion

Increase total delay of vehicles in the intersection

Personal gain

Minimize attacker’s travel time (at the cost of others’) This work

slide-21
SLIDE 21

Analysis approach overview

21

Analysis of Attack input data flow

Data spoofing strategies Traffic snapshots from simulator Congestion creation vuln. Congestion creation exploit

Exploit construction

Dynamic analysis

Spoofing

  • ption enum

Increased delay calc Spoofing w/ high delay inc Source code

slide-22
SLIDE 22

Analysis result summary

22

Analysis of Attack input data flow

Data spoofing strategies Traffic snapshots from simulator Congestion creation vuln. Congestion creation exploit

Exploit construction

Dynamic analysis

Spoofing

  • ption enum

Increased delay calc Spoofing w/ high delay inc Source code

2 distinct types of algorithm-level vulnerabilities:

One single attack vehicle can greatly manipulate traffic control!

slide-23
SLIDE 23

I-SIG system

23

I-SIG

1 2 3 4 5 6 7 8

slide-24
SLIDE 24

COP (Controlled Optimization of Phases)

24

I-SIG

5

5 sec

5

Signal plan (green light length & order) with lowest total delay Input: All vehicles’ location & speed Output:

3 7

2

3 Delay = 15 Delay = 0 Delay = 0

1 1: 5 sec 2: 3 sec 1: 7 sec

(total delay: 15 sec) Dynamic programming

5 3 1

slide-25
SLIDE 25

COP (Controlled Optimization of Phases)

25

I-SIG

5 5 3 7

2

3 Delay = 15 Delay = 0 Delay = 0

1

5 3 1

Data from one single vehicle: Very hard to affect signal plan

+3×n

  • Commonly, 1 vehicle vs > 25 vehicles’

delay in 5 conflicting lanes

  • Can’t change even 1 sec

+n

1

+n

slide-26
SLIDE 26

5 3 7

Vuln #1: Last vehicle advantage

  • Attack: Spoof to arrive as late as possible to increase the

delay of queuing vehicles in other lanes

26

9 9 Delay = 15 Delay = 0 Delay = 0

5 3 1 +12

I-SIG

…

40

+105

5

1 2

slide-27
SLIDE 27

Cause: Effectiveness & timeliness trade-off

  • COP on RSU = 4-5 sec

decision time < 3 sec

  • To meet timeliness requirement, customize COP to limit the

# of servings per lane

– By default, only serve each lane once

27

9 Delay = 15 Delay = 0 Delay = 0

5 3 1 +12

I-SIG

Timeliness Security Effectiveness Sub-optimal COP also good 5 3 7 9 5 Unexpectedly exposed vuln.

Sub-optimal COP

1 2

RSU = Road-Side Unit

slide-28
SLIDE 28

Vuln #2: Curse of transition period

  • I-SIG has 2 operation modes based on PR:

– PR ≥ 95%, full deployment: Directly run COP – PR < 95%, transition: COP becomes ineffective, use an unequipped vehicle estimation algorithm as pre-processing step

28

PR ≥ 95% Unequipped vehicle estimation

COP algorithm

Yes (full deployment period) No (transition period)

PR = Penetration Rate

slide-29
SLIDE 29

Unequipped vehicle estimation algorithm

29

PR ≥ 95% Unequipped vehicle estimation

COP algorithm

Yes (full deployment period) No (transition period)

PR = Penetration Rate

Queuing region Slow-down region Free flow region Vulnerable

slide-30
SLIDE 30

Vulnerable queue estimation

  • Data from one single attack vehicle can add 30-50 “ghost”

vehicles to COP input

  • Dramatically increase length of (wasted) green light

30

Spoof the vehicle location!

  • Est. queue length = 3
  • Est. queue length = 7
slide-31
SLIDE 31

Attack video demo

  • Demo time!

– https://www.youtube.com/watch?v=3iV1sAxPuL0

31

slide-32
SLIDE 32

Recent interest: Autonomy software security in smart transportation

32

Connected Vehicle (CV) Autonomous Vehicle (AV)

[ISOC NDSS’18] First software security analysis of a CV-based transportation system [ACM CCS’19] First software security analysis of LiDAR-based AV perception

slide-33
SLIDE 33

Recent interest: Autonomy software security in smart transportation

33

Connected Vehicle (CV) Autonomous Vehicle (AV)

[ISOC NDSS’18] First software security analysis of a CV-based transportation system [ACM CCS’19] First software security analysis of LiDAR-based AV perception

slide-34
SLIDE 34

Background: Autonomous Vehicle technology

  • Equip vehicles with various types of sensors to

enable self driving

34

slide-35
SLIDE 35

Goal: First security analysis of AV autonomy software

  • New attack surface: Sensors

– Key input channel for critical control decisions – Public channel shared with potential adversaries

  • Fundamentally unavoidable attack surface

35

slide-36
SLIDE 36
  • Camera/LiDAR/RADAR:

– Spoofing attack: inject spoofed obstacles -> emergency brake, rear-end collision etc.

Background: AV autonomy software & possible sensor attacks

36

slide-37
SLIDE 37
  • Camera/LiDAR/RADAR:

– DoS attack: prevent victim from performing

  • bject detection -> collide into a front vehicle

Background: AV autonomy software & possible sensor attacks

37

slide-38
SLIDE 38
  • GPS:

– Spoofing attack: Make victim deviate from the lane

  • > crash into cars in the wrong way or adjacent lanes

Background: AV autonomy software & possible sensor attacks

38

slide-39
SLIDE 39
  • GPS:

– DoS attack: Victim unable to localize itself -> deviate from lane -> crash to cars in wrong way or adj. lanes

Background: AV autonomy software & possible sensor attacks

39

slide-40
SLIDE 40

Goal: First security analysis of AV autonomy software

  • New attack surface: Sensors

– Key input channel for critical control decisions – Public channel shared with potential adversaries

  • Fundamentally unavoidable attack surface!
  • LiDAR

40

slide-41
SLIDE 41

Background: LiDAR basics

41

slide-42
SLIDE 42

Background: LiDAR attacks

  • Known attack: LiDAR spoofing1

– Shoot laser to LiDAR to inject points

42

1 Shin et al.@CHES’17

How to use this to attack AV software control logic?

slide-43
SLIDE 43

First security analysis of LiDAR-based perception in AV

  • Target: Baidu Apollo AV software system

– Production-grade system, drive some buses in China already – Open sourced (“Android in AV ecosystem”) – Partner with 100+ car companies, including BMW, Ford, etc.

  • Attack: LiDAR spoofing attack from road-side laser

shooting devices to create fake objects

– Trigger undesired control operations, e.g., emergency brake

43

Set up road-side device to shoot laser

Fake

  • bject
slide-44
SLIDE 44

LiDAR input workflow in Apollo

44

ROI filter Data aggregation Deep learning model

Point cloud data Objectness

slide-45
SLIDE 45

LiDAR input workflow with attack

45

ROI filter Data aggregation Deep learning model

Point cloud data Spoofed data points from LiDAR spoofing Objectness

slide-46
SLIDE 46

LiDAR input workflow with attack

46

ROI filter Data aggregation Deep learning model

Point cloud data Objectness Data trace of LiDAR spoofing Attack parameters:

  • Rotation
  • Scale
  • Height

Attack data synthesis

slide-47
SLIDE 47

Analysis approach

47

ROI filter Data aggregation Deep learning model

Point cloud data Objectness Data trace of LiDAR spoofing Attack parameters:

  • Rotation
  • Scale
  • Height

Attack data synthesis

Gradient descent

Input: Math function Increase Change

slide-48
SLIDE 48

Analysis approach

48

ROI filter Data aggregation Deep learning model

Point cloud data Objectness Data trace of LiDAR spoofing Attack parameters:

  • Rotation
  • Scale
  • Height

Attack data synthesis

Gradient descent

Input: Math function Increase Change Math function for pre-processing steps

Model

slide-49
SLIDE 49

Analysis results

  • Successfully find

attack input that can inject fake object!

49

slide-50
SLIDE 50

Security implication: Emergency brake attack

50

  • Cause AV to decrease speed from 43km/h to

0 km/h within 1 sec!

slide-51
SLIDE 51

Security implication: Car “freezing” attack

51

  • “Freeze” an AV at an intersection forever!
slide-52
SLIDE 52

Conclusion

  • Initiated the first research efforts to perform security analysis
  • f autonomy software in CV/AV systems
  • Discovered new attacks, analyzed root causes, and

demonstrated security & safety implications

  • Only the beginning of CV/AV software security research

– Initiated the ACM AutoSec workshop to build community – Interested in joining? Fill this form: https://forms.gle/S7QzGkVMTcLzFvcT8

Contact:

Qi Alfred Chen Computer Science, UC Irvine Email: alfchen@uci.edu Homepage: https://www.ics.uci.edu/~alfchen/