1. Random Number Generation – Lava Lamps, Clouds and the IoT
OWASP Meetup
Richard Moulds - Vice President Strategy, Whitewood
January 31st 2017

2. Cryptography – the basis of digital security
Digital Certificates
(authentication)
Encryption
(data confidentiality)
Digital signatures
(integrity and non-repudiation)
Protect data at rest
Strong authentication
Code signing
Secure time
Secure communications
Mobile payments
Secure archives

3. Crypto is all about secrets
Outsider the ‘Perimeter’
Inside the ‘Perimeter’
Encryption
Decryption
Encrypted Data
Data
Math
Math
Data
Network traffic
Backup media
Forensic requests
Portable media
Cloud storage
File shares
keys
keys
Outsiders can only try to guess the keys
Insiders focus on stealing the keys
Two issues
First - more and more stuff is outside the perimeter
Second (next side) – keys are getting easier to guess

4. All crypto security starts with random numbers
What if there was a threat that was almost impossible to detect, is barely addressed by any security product and is getting worse every day?
Security assumptions rely on keys being truly random - when patterns emerge (or are engineered), keys get predictable and crypto is weakened

5. Hidden vulnerabilities and backdoors of choice

6. Testing for randomness
1.0
Single die
Two dice
Loaded dice
Probabilities of outcomes
Measuring uniformity and lack of bias is a good start…
It’s easy when you know where the numbers come from – but what if you don’t

7. Proving unpredictability is more tricky
What data looks the most unpredictable? 7 𝜋
For crypto we also need unpredictability, imperturbability, secrecy and reliability all of which requires knowledge of the source of randomness, not just statistical analysis of the output
It’s easy when you know where the numbers come from – but what if you don’t

8. Finally we have a standard (nearly)
“Specifying an entropy source is a complicated matter. This is partly due to confusion in the meaning of entropy, and partly due to the fact that, while other parts of an RBG design are strictly algorithmic, entropy sources depend on physical processes that may vary from one instance of a source to another”.
Pick up on the second quote
Source – Recommendation for the Entropy Sources Used for Random Bit Generation (SP800-90B 2nd draft) – NIST January 2016

9. RANDOM NUMBER GENERATOR
Why so complicated?
RANDOM NUMBER GENERATOR
Most random numbers come from the Operating System
But software doesn’t act randomly

10. Entropy - a long standing issue
“Anyone who considers arithmetical methods of producing random digits is, of course, in a state of sin.” (J. von Neumann, 1951)

11. Pseudo-random numbers – an oxymoron?
Operating System
Random Numbers
Random Seeds
Entropy Source
Pseudo-random number generator
Crypto Application
Shuffling the deck
Dealing the deck

12. Where does entropy come from?
App1
App2
App3
Operating System
Host System
Random Numbers
Pseudo-random number generator
Hardware
CPU Timing
Network Timing
Hard Drive Timing
Entropy
Mouse Clicks
Camera
Antenna
Local Environment
Microphone
Keyboards

13. Pseudo-random number generator
But in a virtual world…
App1
App2
App3
Hardware
CPU Timing
Network Timing
Mouse Clicks
Camera
Antenna
Local Environment
Microphone
Keyboards
Host System
Hard Drive Timing
Random Numbers
Hypervisor
Operating System
Pseudo-random number generator

14. Random number generators in Linux
dev/urandom
dev/random
Delivers random numbers only if sufficient entropy has been captured - otherwise it stops
Delivers random numbers irrespective of how much entropy has been captured

15. Entropy sources in Linux
Interrupt Entropy Pool (1024 bits)
Main Entropy Pool (4096 bits)
Interrupt events
/dev/urandom
PRNG
Disk events,
keyboard clicks and mouse movements
/dev/random
PRNG

16. Interrupts (add_interrupt_randomness)
Kernel IRQ handler adds data from each interrupt into the Interrupt Pool
One Interrupt pool per CPU to eliminate contention
Cycle counter XOR kernel timer
IRQ number
Instruction pointer at the time the interrupt is received
Cycle Count & Kernel Timer
IRQ
Instruction Pointer
4 bytes
4 bytes
8 bytes

17. Interrupts (add_interrupt_randomness)
Kernel IRQ handler adds data from each interrupt into the Interrupt Pool
One Interrupt pool per CPU to eliminate contention
Cycle Count & Kernel Timer
IRQ
Instruction Pointer
4 bytes
4 bytes
8 bytes
Cycles
Kernel
IRQ
Instruction Pointer 14

18. Disk (add_timer_randomness)
Disk events are funneled through timer randomness
One Interrupt pool per CPU to eliminate contention
Kernel Timer
Cycle Counter
Device id (disk_devt)
4 bytes
4 bytes
8 bytes
Kernel Timer
Cycles
Device ID

19. Enhancing system entropy
Goal: generate true random numbers from a PRNG
Existing applications
Entropy is always additive
‘True’ random numbers
Supplementary entropy source(s)
Operating
System
PRNG
e.g. /dev/random
Existing system entropy

20. Supplementary sources of entropy
4 general ways to improve entropy beyond the basic kernel:
Software daemons to extract better timing related entropy:
HAVEGED – (
CPU Jitter RNG (
Entropy extraction from peripheral devices (mics and cameras)
audio-entropyd & video-entropyd - (
Local hardware based entropy sources
Embedded CPU feature (RDRAND), USB devices, PCI cards, etc.
Wide range of noise sources – electrical, meta-stable circuits, quantum
Wiki search - “comparison of hardware random number generators”
Network based sources – “Entropy as a Service”
(random numbers rather than entropy)
NIST (coming soon?)
Whitewood (

21. Comparison of supplementary entropy sources
Jitter Daemons
Noisy sensors
Hardware RNGs
Entropy as a Service
Primary focus
Application specific
Individual machine
Distributed systems
Scalability
Medium
Poor
Low - High
High
Maturity
Open source
Niche
Mature
Emerging
Assurance
Low
High*
Visibility
Control
Medium (black-box)
High (private service)
Cost
Free
Sensor?
$0 - $10k
Amortized
In a Nutshell
Band Aid
For the Hobbyists
“No one likes hardware”
Infrastructure of the future?
* - when new NIST standard is finalized

22. Whitewood Entropy Engine
Quantum Random Number Generator (QRNG)
Generates random numbers using the quantum properties of light
Quantum noise source is 100% unpredictable - independent of all external factors
Delivers extremely high performance
Output data rate of 350Mbit/s
Deployed as local source or network service
Designed to comply with NIST B/C
Based on 20 years research at Los Alamos
Entropy Engine PCIe card

23. Summary
Encryption and cryptography are the basis of trust and security in the digital world
Random numbers are critical for security but are often poorly understood and managed
Random number generators are a point of attack and vulnerability – potentially an invisible one
Modern application environments present entropy challenges – cloud, appliance, mobile, browser, IoT
Proving the operation and quality of entropy sources and random number generators is difficult
New standards such as NIST will help
Random number generation should be a critical component of your key management strategy and datacenter infrastructure

24. Demo at www.whitewoodencryption.com/netrandom-demo
Thank you
Demo at