Download Hardware YOU Can Trust

Hardware YOU Can
Trust
Benedikt Stockebrand
Stepladder IT Training+Consulting GmbH
c 2015 Benedikt Stockebrand
Copyright March 18, 2015
Troopers 15
Heidelberg, Germany
1/49
Hardware YOU Can
Audit and Then
Trust
Benedikt Stockebrand
Stepladder IT Training+Consulting GmbH
c 2015 Benedikt Stockebrand
Copyright March 18, 2015
Troopers 15
Heidelberg, Germany
2/49
About Me
• Diplom-Informatiker (Uni Dortmund)
• Specialized in IT operations, TCP/IP networks, Unix
• Even more specialized on IPv6 since mid 2003
• Co-author of an IPv6 related security study
for the BSI (Federal Office for IT Security)
• Actively working on microcontrollers and
hardware random number generators since 2012 (or so)
c 2015 Benedikt Stockebrand
Copyright 3/49
Part I
The Current Situation
c 2015 Benedikt Stockebrand
Copyright 4/49
So-called “IT Security”
c 2015 Benedikt Stockebrand
Copyright The Current Situation
5/49
So-called “IT Security”
The Current Situation
• Do we have to prove IT systems to be insecure?
c 2015 Benedikt Stockebrand
Copyright 5/49
So-called “IT Security”
The Current Situation
• Do we have to prove IT systems to be insecure?
• . . . or should the vendor demonstrate their security instead?
c 2015 Benedikt Stockebrand
Copyright 5/49
So-called “IT Security”
The Current Situation
• Do we have to prove IT systems to be insecure?
• . . . or should the vendor demonstrate their security instead?
• For more on this:
“The Limits of Cryptography”, EasterHegg 2014
http://www.youtube.com/watch?v=7bTaKSZQKhc
c 2015 Benedikt Stockebrand
Copyright 5/49
“Secure Hardware”
The Current Situation
• Do we blindly want to trust all the
• designers/developers
• vendors
• manufacturers
• component suppliers
• logistics operators
• distributors
• dealers
• customs authorities
• investigative authorities
• so-called “intelligence” agencies
• standard issuing bodies
• . . . and whoever else
involved?
c 2015 Benedikt Stockebrand
Copyright 6/49
“Tamper Protection”?
The Current Situation
• In some cases, this is imporant
c 2015 Benedikt Stockebrand
Copyright 7/49
“Tamper Protection”?
The Current Situation
• In some cases, this is imporant
• But in other cases this is
c 2015 Benedikt Stockebrand
Copyright 7/49
“Tamper Protection”?
The Current Situation
• In some cases, this is imporant
• But in other cases this is
• unnecessary
c 2015 Benedikt Stockebrand
Copyright 7/49
“Tamper Protection”?
The Current Situation
• In some cases, this is imporant
• But in other cases this is
• unnecessary
• exceedingly difficult
c 2015 Benedikt Stockebrand
Copyright 7/49
“Tamper Protection”?
The Current Situation
• In some cases, this is imporant
• But in other cases this is
• unnecessary
• exceedingly difficult
• or a convenient excuse to peddle snake oil.
c 2015 Benedikt Stockebrand
Copyright 7/49
Auditability I
The Current Situation
• Can I audit a TPM module or the crypto features of a CPU. . .
• . . . without destroying it?
• . . . with affordable and available tools?
• . . . with available know-how?
• . . . and with a reasonable amount of effort?
c 2015 Benedikt Stockebrand
Copyright 8/49
Auditability II
The Current Situation
• Is auditability possible?
c 2015 Benedikt Stockebrand
Copyright 9/49
Auditability II
The Current Situation
• Is auditability possible?
• Not with attackers, who
• have unlimited resources
• have unlimited technical skills
• have unlimited insider knowledge
• have unlimited political and judicial backing
• nonchalantly ignore applicable laws
(not to talk about “decency”)
• have support from the manufacturers
• don’t make mistakes
c 2015 Benedikt Stockebrand
Copyright 9/49
Auditability II
The Current Situation
• Is auditability possible?
• Not with attackers, who
• have unlimited resources
• have unlimited technical skills
• have unlimited insider knowledge
• have unlimited political and judicial backing
• nonchalantly ignore applicable laws
(not to talk about “decency”)
• have support from the manufacturers
• don’t make mistakes
• But what about real world attackers?
c 2015 Benedikt Stockebrand
Copyright 9/49
Risk Assessment of Attacks
• Risk of discovery
• Personal risk for attacker
c 2015 Benedikt Stockebrand
Copyright 10/49
The Current Situation
Risk Assessment of Attacks
• Risk of discovery
• Personal risk for attacker
• Economic risk for attacker
c 2015 Benedikt Stockebrand
Copyright 10/49
The Current Situation
Scope of Attacks
The Current Situation
• Targeted attacks
• Untargeted, large scale attacks
c 2015 Benedikt Stockebrand
Copyright 11/49
Attack Types
The Current Situation
• Passive snooping
• Active manipulation/injection of fake data
c 2015 Benedikt Stockebrand
Copyright 12/49
Attack Types
The Current Situation
• Passive snooping
• Active manipulation/injection of fake data
• Installation of backdoors in operational devices
c 2015 Benedikt Stockebrand
Copyright 12/49
Attack Types
The Current Situation
• Passive snooping
• Active manipulation/injection of fake data
• Installation of backdoors in operational devices
• Replacement of operational devices
c 2015 Benedikt Stockebrand
Copyright 12/49
Attack Types
The Current Situation
• Passive snooping
• Active manipulation/injection of fake data
• Installation of backdoors in operational devices
• Replacement of operational devices
• Targeted distribution of manipulated devices
c 2015 Benedikt Stockebrand
Copyright 12/49
Attack Types
The Current Situation
• Passive snooping
• Active manipulation/injection of fake data
• Installation of backdoors in operational devices
• Replacement of operational devices
• Targeted distribution of manipulated devices
• Manipulation of entire product series or product families
c 2015 Benedikt Stockebrand
Copyright 12/49
Attack Types
The Current Situation
• Passive snooping
• Active manipulation/injection of fake data
• Installation of backdoors in operational devices
• Replacement of operational devices
• Targeted distribution of manipulated devices
• Manipulation of entire product series or product families
• Manipulation of standards and specifications
c 2015 Benedikt Stockebrand
Copyright 12/49
Attack Types
The Current Situation
• Passive snooping
• Active manipulation/injection of fake data
• Installation of backdoors in operational devices
• Replacement of operational devices
• Targeted distribution of manipulated devices
• Manipulation of entire product series or product families
• Manipulation of standards and specifications
• Attacks at the political level
c 2015 Benedikt Stockebrand
Copyright 12/49
c 2015 Benedikt Stockebrand
Copyright Part II
Fighting Back
13/49
Defense Strategies I
Fighting Back
• Technical
c 2015 Benedikt Stockebrand
Copyright 14/49
Defense Strategies I
Fighting Back
• Technical
• Economic
c 2015 Benedikt Stockebrand
Copyright 14/49
Defense Strategies I
Fighting Back
• Technical
• Economic
• Political
c 2015 Benedikt Stockebrand
Copyright 14/49
Defense Strategies II
Fighting Back
• Diversity
• Auditability
• Modularity
c 2015 Benedikt Stockebrand
Copyright 15/49
Development Criteria
Fighting Back
• Designs must be
• easy to understand
• easy to modify
• easy to reproduce (and build yourself)
• Development tools must be
• auditable
• generally available
• affordable
• available in multiple implementations
c 2015 Benedikt Stockebrand
Copyright 16/49
Usable Hardware Components
Components should be
• cheap
• mass produced for multiple purposes
• available from multiple manufacturers
• simple (i.e. no highly integrated ICs if possible)
• widely available
• easily replaceable by similar components
• easy to use for custom builds
c 2015 Benedikt Stockebrand
Copyright 17/49
Fighting Back
PCB Technology
Fighting Back
• Through-hole (THT) only
• Surface mount (SMT) for home assembly
• Surface mount for industrial assembly
c 2015 Benedikt Stockebrand
Copyright 18/49
Part III
A Random Self-Experiment
c 2015 Benedikt Stockebrand
Copyright 19/49
Choosing the External Interface
• Raw analog output
• UART (5V)
• RS-232
• USB
• PCIe
c 2015 Benedikt Stockebrand
Copyright 20/49
A Random Self-Experiment
Choosing the Microcontroller and
Interface
• Atmel ATtiny2313 and FTDI FT232RL
• Various MSP430 and FTDI FT232RL
• PIC 16F 1454/1455/1459
• PIC 18F 13K50/14K50
• Various Atmel ATxmega with on-chip USB
• Various STM32
c 2015 Benedikt Stockebrand
Copyright 21/49
A Random Self-Experiment
FTDIgate
A Random Self-Experiment
• FT232’s can be configured via USB interface
• In September 2014, FTDI released a driver update. . .
c 2015 Benedikt Stockebrand
Copyright 22/49
FTDIgate
A Random Self-Experiment
• FT232’s can be configured via USB interface
• In September 2014, FTDI released a driver update. . .
• . . . which bricked fake FTDI chips
• (and no warning given)
c 2015 Benedikt Stockebrand
Copyright 22/49
FTDIgate
A Random Self-Experiment
• FT232’s can be configured via USB interface
• In September 2014, FTDI released a driver update. . .
• . . . which bricked fake FTDI chips
• (and no warning given)
• That chip must go
c 2015 Benedikt Stockebrand
Copyright 22/49
Choosing the Noise Source
• Radioactive decay
• Radio static
• Noise from blackened CCD chip
• Thermal noise in a resistor
• Ring oscillator
• Microphone at Kindergarten playground
• ...
c 2015 Benedikt Stockebrand
Copyright 23/49
A Random Self-Experiment
Choosing the Noise Source
• Radioactive decay
• Radio static
• Noise from blackened CCD chip
• Thermal noise in a resistor
• Ring oscillator
• Microphone at Kindergarten playground
• ...
• Avalanche effect in “Zener” diodes
c 2015 Benedikt Stockebrand
Copyright 23/49
A Random Self-Experiment
Generating 12V
A Random Self-Experiment
• Fredrik Thulin’s choice: Charge pump
• Depends on input voltage
• Doesn’t need inductances
c 2015 Benedikt Stockebrand
Copyright 24/49
Generating 12V
A Random Self-Experiment
• Fredrik Thulin’s choice: Charge pump
• Depends on input voltage
• Doesn’t need inductances
• My choice: Step-up converter (MC34063)
• Needs inductances
• Ultra cheap
• Huge variety of vendors
• Widely known
• General purpose chip
c 2015 Benedikt Stockebrand
Copyright 24/49
Block Schematic
A Random Self-Experiment
Microcontroller
UART↔USB
Interface
USB
Connector
Amplifier
Noise
Generator
Step-Up
Converter
c 2015 Benedikt Stockebrand
Copyright 25/49
Noise Specifications
A Random Self-Experiment
• Noise is undesirable. . .
c 2015 Benedikt Stockebrand
Copyright 26/49
Noise Specifications
A Random Self-Experiment
• Noise is undesirable. . .
• . . . and generally specified as upper limit only
c 2015 Benedikt Stockebrand
Copyright 26/49
Noise Specifications
A Random Self-Experiment
• Noise is undesirable. . .
• . . . and generally specified as upper limit only
• This is a fundamental problem.
c 2015 Benedikt Stockebrand
Copyright 26/49
“Magic Zener Diodes”
c 2015 Benedikt Stockebrand
Copyright A Random Self-Experiment
27/49
“Magic Zener Diodes”
A Random Self-Experiment
Source: http://www.conrad.de/ce/de/product/179001
c 2015 Benedikt Stockebrand
Copyright 27/49
Getting Reliable Noise
A Random Self-Experiment
• More magic?
• BE junction in bipolar junction transistors (BJTs)
c 2015 Benedikt Stockebrand
Copyright 28/49
Getting Reliable Noise
A Random Self-Experiment
• More magic?
• BE junction in bipolar junction transistors (BJTs)
• This is still a fundamental problem.
c 2015 Benedikt Stockebrand
Copyright 28/49
Building for Auditability
A Random Self-Experiment
• Minimized circuit design
• Test pins
• Auditable PCB layout
• DIY compatibility
c 2015 Benedikt Stockebrand
Copyright 29/49
Minimizing the Circuit Design
A Random Self-Experiment
2
1
TP4
10uH
L3
10uH
D1.A
R5
6.8V
100nF
D2
R7
D3
TP8
OUTPUT
T2.B
Q1
BC547C
Q2
BC547C
C5
GND
GND
Amplifier Stage
Generator Core
c 2015 Benedikt Stockebrand
Copyright GND
GND
GND
30/49
1 JP6
2
27K
BAT85
TP3
T1.B
1k
C7
TP5
RAW_OUT
TP6
100uF
R8
+
470
C6
1 JP4
2
1k
TP2
R6
+
C4
100uF
TP7
VCC_AMP_STABLE
+HIGH_V_STABLE
100nF
L2
1 JP3
2
JP5
My Very First Real PCB
c 2015 Benedikt Stockebrand
Copyright A Random Self-Experiment
31/49
Analog Output I
c 2015 Benedikt Stockebrand
Copyright A Random Self-Experiment
32/49
Analog Output II
c 2015 Benedikt Stockebrand
Copyright A Random Self-Experiment
33/49
Analog Output III
c 2015 Benedikt Stockebrand
Copyright A Random Self-Experiment
34/49
Analog Output IV
c 2015 Benedikt Stockebrand
Copyright A Random Self-Experiment
35/49
Amplified Output I
c 2015 Benedikt Stockebrand
Copyright A Random Self-Experiment
36/49
Amplified Output II
c 2015 Benedikt Stockebrand
Copyright A Random Self-Experiment
37/49
Amplified Output III
c 2015 Benedikt Stockebrand
Copyright A Random Self-Experiment
38/49
Amplified Output IV
c 2015 Benedikt Stockebrand
Copyright A Random Self-Experiment
39/49
Amplified Output V
c 2015 Benedikt Stockebrand
Copyright A Random Self-Experiment
40/49
Digitizing Noise
A Random Self-Experiment
• Trivial approach: XOR n readings
• Fast
• Constant speed
• Not adapting to quality of analog source
• Not failsafe
c 2015 Benedikt Stockebrand
Copyright 41/49
Digitizing Noise
A Random Self-Experiment
• Trivial approach: XOR n readings
• Fast
• Constant speed
• Not adapting to quality of analog source
• Not failsafe
• Measure (LSB of) time between rising edges
• Slightly slower
• Variable speed
• Adapts to quality of analog source
• Failsafe behaviour
• Automatically removing correlation
c 2015 Benedikt Stockebrand
Copyright 41/49
From Noise to Entropy
A Random Self-Experiment
• Correlation
• Removed via rising edge algorithm
• Analog circuit is too simple to store much information,
⇒ No correlation between bits possible
c 2015 Benedikt Stockebrand
Copyright 42/49
From Noise to Entropy
A Random Self-Experiment
• Correlation
• Removed via rising edge algorithm
• Analog circuit is too simple to store much information,
⇒ No correlation between bits possible
• (Bitwise) bias
• Follow up with a von Neumann extractor:
• Read two bits
• If the same, discard both
• Else return the first
• Rinse and repeat as needed
c 2015 Benedikt Stockebrand
Copyright 42/49
Why the “ARRGH” board?
A Random Self-Experiment
• It’s the Auditable Real Random Number Generator Hardware
• It has nothing to do with a Makefile bug that made me
generate four weeks worth of test data using the XOR
extraction on a less-than-magic Zener diode. . .
c 2015 Benedikt Stockebrand
Copyright 43/49
c 2015 Benedikt Stockebrand
Copyright Part IV
Preliminary Finale
44/49
Conclusions
Preliminary Finale
• ICs suck (especially from FTDI)
• THT components suck
• SMD/SMT-only components suck
• USB sucks even worse
c 2015 Benedikt Stockebrand
Copyright 45/49
Conclusions
Preliminary Finale
• ICs suck (especially from FTDI)
• THT components suck
• SMD/SMT-only components suck
• USB sucks even worse
• Auditable hardware is difficult
• . . . but apparently possible
c 2015 Benedikt Stockebrand
Copyright 45/49
Conclusions
Preliminary Finale
• ICs suck (especially from FTDI)
• THT components suck
• SMD/SMT-only components suck
• USB sucks even worse
• Auditable hardware is difficult
• . . . but apparently possible
• Complexity is a problem, not a solution
c 2015 Benedikt Stockebrand
Copyright 45/49
Conclusions
Preliminary Finale
• ICs suck (especially from FTDI)
• THT components suck
• SMD/SMT-only components suck
• USB sucks even worse
• Auditable hardware is difficult
• . . . but apparently possible
• Complexity is a problem, not a solution
• Randomness looks really simple
• . . . until you really take a look
c 2015 Benedikt Stockebrand
Copyright 45/49
Outlook: Testing
Preliminary Finale
• There are no proper test methods
• . . . let alone ready-to-use tools
c 2015 Benedikt Stockebrand
Copyright 46/49
Outlook: Testing
Preliminary Finale
• There are no proper test methods
• . . . let alone ready-to-use tools
• Testing for pseudo randomness is much better known
• . . . but largely unrelated
c 2015 Benedikt Stockebrand
Copyright 46/49
Outlook: Testing
Preliminary Finale
• There are no proper test methods
• . . . let alone ready-to-use tools
• Testing for pseudo randomness is much better known
• . . . but largely unrelated
• Testing for randomness is “the wrong way ’round”
• Best starting point is still Knuth’s ACP Vol. 2
c 2015 Benedikt Stockebrand
Copyright 46/49
Outlook: Testing
Preliminary Finale
• There are no proper test methods
• . . . let alone ready-to-use tools
• Testing for pseudo randomness is much better known
• . . . but largely unrelated
• Testing for randomness is “the wrong way ’round”
• Best starting point is still Knuth’s ACP Vol. 2
• The NIST screwed this up again. Big time.
c 2015 Benedikt Stockebrand
Copyright 46/49
Outlook: Testing
Preliminary Finale
• There are no proper test methods
• . . . let alone ready-to-use tools
• Testing for pseudo randomness is much better known
• . . . but largely unrelated
• Testing for randomness is “the wrong way ’round”
• Best starting point is still Knuth’s ACP Vol. 2
• The NIST screwed this up again. Big time.
• But this is enough for another year of research. . .
• . . . and hopefully another talk
c 2015 Benedikt Stockebrand
Copyright 46/49
c 2015 Benedikt Stockebrand
Copyright Part V
Appendix
47/49
Resources
Appendix
The Limits of Cryptography
EasterHegg 2014, Stuttgart
http://www.youtube.com/watch?v=7bTaKSZQKhc
BIVBlog: Benedikt’s IT Video Blog
http://www.stepladder-it.com/bivblog/
Video Blog on IT in general and crypto hardware (and IPv6) in
particular
The Cryptech project
https://cryptech.is/
c 2015 Benedikt Stockebrand
Copyright 48/49
Appendix
Contact Information
c 2015 Benedikt Stockebrand
Copyright Stepladder IT
Training+Consulting GmbH
Benedikt Stockebrand
Fichardstr. 38
D-60322 Frankfurt/Main
[email protected]
Webseiten:
http://www.stepladder-it.com/
http://www.benedikt-stockebrand.de/
Video Blog:
http://www.stepladder-it.com/bivblog/
49/49