Bug: challenge_response.c uses predictable rand()/srand() for nonce generation

[508704820]

Bug: C challenge-response uses rand()/srand() — predictable nonce

Severity: High (Security)

File: src/anti_spoof/challenge_response.c, generate_challenge() (line 300) and main() (line 519)

Description:
The C implementation uses rand() seeded with srand(time(NULL) ^ read_timebase()) to generate challenge nonces. Both rand() (Mersenne Twister/LCG) and the seed are predictable:

1. Line 300: c.nonce[i] = (unsigned char)(rand() ^ (c.timestamp >> (i % 8))); — XOR with known timestamp doesn't help if rand() output is predictable
2. Line 519: srand((unsigned int)time(NULL) ^ (unsigned int)read_timebase()) — seed is guessable

Impact:

• Challenge nonces can be predicted by an attacker who knows approximate timestamp
• Anti-spoof challenges can be pre-computed, defeating the entire purpose
• Consistent with the Python-side PRNG bugs already reported ([SECURITY] Deep entropy challenge uses predictable PRNG - Mersenne Twister for challenge generation #4640, [SECURITY] Proof of Antiquity select_proof() uses predictable random.choice/uniform for weighted selection #4643)

Expected Fix:
Use getrandom() (Linux) or SecRandomCopyBytes() (macOS) for nonce generation. On PowerPC, use the timebase register directly as entropy source.