Why rand() is Predictable

Most languages' built-in random functions are NOT cryptographically secure:

// C
srand(time(NULL));
int x = rand();   // predictable from time

// Python
import random
random.random()  // Mersenne Twister; predictable from 624 outputs

These use pseudo-random number generators (PRNGs) — deterministic algorithms producing seemingly-random output from a SEED. Given the seed, all output is reproducible.

Example: Mersenne Twister (Python's random):

• 19937-bit internal state
• Once an attacker observes 624 32-bit outputs, they can SOLVE for the state and predict ALL future output

Why is this OK for some uses?

• Simulations (need reproducibility)
• Games (just feel-random)
• Hash table seed (within trusted process)

NOT OK for:

• Crypto keys
• Session tokens
• Password reset tokens
• IV/nonce for encryption
• ANY security-relevant value

For security: use CSPRNG (Cryptographically Secure PRNG). The OS provides one:

• Linux: /dev/urandom or getrandom()
• macOS: /dev/urandom or arc4random
• Windows: BCryptGenRandom
• Java: SecureRandom
• Python: secrets module
• Node: crypto.randomBytes()