This document discusses security issues related to accessing and controlling vehicles via OBD-II ports, drawing comparisons to struggles securing industrial control systems. It notes that accessing these systems often means compromising them, as protocols were designed without security. While an analysis of a Progressive Snapshot dongle found no security precautions, lessons from securing critical infrastructure suggest restricting access and implementing least privilege. The document advocates learning from past ICS mistakes to develop secure vehicle protocols and modules.

1. Attacking and Defending Autos
Via OBD-II
Dale Peterson, Digital Bond, Inc.
peterson@digitalbond.com
@digitalbond

2. Industrial Control Systems

3. Monitor and Control A Process
• Sensors measure temperature, current, speed,
flow, …
• Actuators turn things on/off, heat/cool, mix,
control flow, …
• Some are controlled by humans/operators
• Others maintain steady state or goal
• Additional safety systems prevent really bad
things from happening
• Sound familiar?

4. Let’s Learn From Industrial Control
System (ICS) Security Struggles
and Save Decades of Insecurity in
the Auto Industry

5. ICS Access = Compromise
• If an attacker can gain access to almost any
SCADA, DCS or other type of ICS
– He can cause components to crash
– He can control the process
– He can change the process (Stuxnet)
– He can sometimes change the Safety System (Stuxnet)

6. ICS Protocols – Insecure By Design

7. Secure By Design
• Product goes through a Security Development
Lifecycle (SDL)
– Microsoft and others provide resources on this
• Threat Model determines required security controls
• Software coding practices
• Fuzz testing
• Independent 3rd Party Testing
• Well understood but takes years & effort

8. Insecure By Design
Insecure By Design is not simply the lack of
Secure By Design. It is much worse!!!
Insecure By Design provides the attacker with
everything he would want as a documented
feature. There is no need to find a bug or
vulnerability.

9. Insecure By Design
“The pro’s don’t bother with
vulnerabilities when attacking ICS.
They use documented features of
protocols and products.”
Ralph Langner
of Stuxnet fame

10. Charlie Miller / Chris Valasek 2014
• Connected to a control system and …
• Were able to monitor and control, no surprise
• Incredibly impressive reverse engineering
• Completely unnecessary, many in this room have
this knowledge
• The Safety question?
– Chuck Yeager or child jumping off roof?
– Marie Curie or blindly mixing chemicals to see what
happens?

11. No Authentication
• No source or data authentication
• Access = Compromise
– Modbus TCP
– EtherNet/IP
– S7, DNP3, PROFINET, CC, HART, OPC Classic
– Proprietary Protocols
– And CANBus

12. Security Perimeter is Key
• Bad guys will succeed if they get in
• Require effective security perimeter to keep
them out
• ICS History
– Connected directly to corporate networks
– Some connected to the Internet
– Anytime, anywhere remote access for in house
support, 3rd party support, convenience or curiosity

13. Access To Insecure By Design Car?
• Access = Compromise, so how can an attacker
access the CANBus? Many ways are being found.
• Corey Thuen in Digital Bond Labs looked at
insurance use of OBD-II port
• Specifically the Progressive
Snapshot dongle
• Could this comms network,
Snapshot device, or
Progressive site be used
to attack cars?

14. ICS Comparison
1. Access = Compromise? ICS and auto
2. Little or no thought about security in embedded
components

15. Search S4 Basecamp Video

16. Snapshot Dongle

17. Analysis Environment

18. Communication – No Security
• No encryption, no authentication, no firmware
signing or other security
• Attacker has unrestricted access to device if he
can get to it over the cellular network or from a
Progressive network

20. Software Quality
• Did not follow even the most basic secure
coding practices
• Lots of strcopy and other banned functions
– Lead to overflows and other security issues
• Lacking input validation, bad behavior when
fuzzed

21. Snapshot Summary
• Little or no thought to security in development
by vendor (Xirgotech) or Progressive … it’s a little
embedded device
• Deployed units can never be trusted
• Likely some improvements has taken place in
backend servers at Progressive which is very
important
– Compromising Progressive Servers could allow access
and control to 2,000,000+ vehicles

22. ICS Comparison
1. Access = Compromise? ICS and auto
2. Little or no thought about security in embedded
components? ICS and auto
3. Push historical data out?

23. GE Power Plant Turbine Monitoring
• 1800+ Turbines in 60+ countries from GE Atlanta
– Useful efficiency and preventive maintenance info

24. GE Power Plant Turbine Monitoring
• Communications from Atlanta includes firewalls,
encryption, two-factor authentication, background
checks, secure facility and more
• From that location in Atlanta they can access (and
control) 1800+ turbines in 60+ countries
• A HUGE target and …
• Control is not necessary, all GE needs is the data
– Sound familiar to insurance companies?

25. Implement Least Privilege
• Only allow what is required
• Some solutions for GE and OBD-II
– Unidirectional (one-way) gateway, data can be sent
from the car out, but no comms allowed to the car
– Limited functionality, allow reads / data access but no
control commands
– Have the equivalent of a historian on the vehicle that
the OBD-II can access
• Evaluate the residual risk

26. ICS and Vehicle Risk Comparison
• End unit compromise
– Much greater for critical infrastructure ICS
– Widespread blackout, lack of gas, environmental
damage, death
– Vehicle compromise could be tragic, but limited
• Greater risk to vehicles is compromise of
concentrated remote access with control
capability
– Benefits are greater in vehicles, but make sure your
risk management includes cyber

27. ICS Comparison
1. Access = Compromise? ICS and auto
2. Little or no thought about security in embedded
components? ICS and auto
3. Push historical data out? ICS today / auto future?
4. Control / Safety segregation

28. Safety Integrated Systems (SIS)
• Prevent really bad things from happening,
automatically, all the time, even when the control
system fails, no human interaction
• Problem: If control system can communicate
with the SIS a cyber attack can compromise both
– Stuxnet Protection system to prevent overpressure
• Problem: Control Systems often want sensor
data from the safety system

29. Increasingly Common Solution
• Install an ICS firewall that only allows read
requests from ICS to SIS
– ICS can’t write to SIS, can’t change logic, load
firmware or do anything but read points
• Solution for cars … restrict communications
between modules
– Focus on protecting modules with most critical control
– Focus on limiting access from modules with remote
access
– Much faster solution than developing secure CAN

30. Develop Secure CAN or
Replace with Secure Protocol
Don’t fall for the “this will take a
decade” like the ICS world did!

31. Digital Bond CANBus Tools
• Digital Bond Labs tools are on GitHub
– https://github.com/digitalbond/
• canbus-utils for analyzing CANBus traffic
• canbus-beaglebone for low-cost testing platform
• canbus-protector includes some proof of concept
fixes discussed in this session

32. Digital Bond
• S4xJapan … November 6th in Tokyo
• S4x16 … January 12-14 in Miami Beach
• Focused control system security company
• Consulting
– Dedicated ICS security team since 2003
• Labs
– Find new vulnerabilities and attack techniques
• Contact: Dale Peterson, peterson@digitalbond.com
@digitalbond on twitter

33. Questions