Property-Based Testing¶

Standard unit tests only verify what the developer expects. Property-Based Testing verifies the system against thousands of randomized adversarial scenarios.

Overview¶

We use the Hypothesis library to prove that the ConsensusGuardian and SafetyGate are resilient under extreme conditions. Instead of testing a single static finding, we generate thousands of randomized finding objects with varying timestamps, nonces, and signatures.

Key Properties Verified¶

1. Temporal Drift Resilience¶

Property: The system must reject any proposal where the timestamp drift exceeds ±60 seconds from "now." - Verification: Hypothesis generates valid and invalid timestamps across a wide range. The guardian correctly identifies and rejects the drifts.

2. Signature Integrity¶

Property: Any modification to the signed finding JSON (even a single character) must result in a signature verification failure. - Verification: We generate valid signatures and then programmatically "tamper" with the payload. The system is proven to block all tested tamper scenarios.

3. Nonce Collision Resistance¶

Property: Reusing a nonce within the TTL window must trigger a mandatory block. - Verification: Hypothesis simulates concurrent requests with duplicate nonces. The TTLCache prevents command-injection via replay attacks.

Why This Matters¶

For autonomous agents, "edge cases" are the norm. By using property-based testing, we achieve: - Verified Resilience: 1,000+ test cases per property covering extreme input ranges. - Adversarial Hardening: Discovery of logic flaws that manual unit tests often miss. - Traceable Assurance: Moving from "it works in simulation" to "demonstrably sound under tested properties."

# Example Hypothesis Test (Simplified)
@given(st.integers(min_value=1, max_value=1000))
def test_quorum_thresholds(n):
    # Proves quorum logic consistency for 'n' agent signatures
    ...