Use cases for random numbers

Requirements for and means to obtain random data for cryptography vary by use. Principal examples include:

Personal computing

• Linux /dev/random, /dev/urandom
• Apple's OS X (160-bit Yarrow SHA-1)
• Other Unix-like systems
• Windows (CryptGenRandom)
• Mobile devices e.g. Smart phone, Tablet
  • Many good sources of randomness (accelerometer, trackpad, camera, radio)
  • Apple's iOS (160-bit Yarrow SHA-1 ??)
  • Android

Server

• Hard drives are a potential source of entropy, but are being replaced by solid state drives (SSDs) which may not provide reliable entropy
• Desktop peripherals (sound input, camera USB port) may be absent
• Virtual machine instances, especially at startup
  • I/O batching and caching may eliminate most entropy from disk file access timing
  • Hypervisor should provide a random data source
  • A CPU RNG instruction would be ideal if trustworthy

Disk-less black boxes

• Routers
• Programmable logic controllers (PLCs)
• Internet of Things
• Systems on a chip (SoC)
  • ARM and SoC vendors do not typically include RNG hardware
• Cost of adding RNG hardware is an obstacle for low price systems.
• First boot and provisioning of security/identity keys requires good TRNG

Crypto appliances

• Commercial TRNGs
• Hardware Security Modules (HSM)
• Data choke
• Randomized instruction set cores on FPGAs
• Authenticating camera & sensors
• Maker movement

DSA nonces

• special case-- arguably a bug, but a widespread problem