Adversary Capabilities
BlindHop inherits the Nym mixnet's threat model. We consider:
| Adversary | Can Observe | Cannot Determine |
|---|
| ISP / Network observer | That you're using Nym | What you're sending, to whom |
| Malicious full node | JSON-RPC requests from exit IP | Who sent the request |
| Malicious exit service | JSON-RPC payload content | Sender's IP or identity |
| Malicious mix node | Encrypted packets passing through | Content, sender, or recipient |
| Colluding nodes (< all) | Partial traffic flow | Full sender-recipient correlation |
Privacy Mode Guarantees
| Threat | None | Fast | Full |
|---|
| Full node learns your IP | ✗ Protected | ✓ Protected | ✓ Protected |
| ISP sees your queries | ✗ Protected | ⚠️ Partially | ✓ Protected |
| Traffic timing analysis | ✗ Protected | ✗ Protected | ✓ Protected |
| Activity detection | ✗ Protected | ✗ Protected | ✓ Protected |
Key Assumptions
- Nym network is honest-majority — at least one mix node per layer is honest
- Sphinx packets are cryptographically sound — x25519 + AES are unbroken
- Cover traffic is sufficient — Loopix parameters provide adequate noise
- Exit service is semi-trusted — it sees content but not sender identity
Limitations
- Full mode provides strong anonymity but with 1-3s latency overhead
- Fast mode hides IP but does not protect against traffic analysis
- None mode provides zero privacy (direct connection)
- The exit service sees plaintext JSON-RPC (but not who sent it)