▶What are the main types of bridge exploits?
Validator collusion (validators steal funds), sybil attacks (attacker controls majority validators), oracle manipulation (bad price data), smart contract bugs (signature verification flawed), and front-running (attacker intercepts transaction). Each requires different mitigations. No bridge is exploit-proof; only risk-managed.
▶What's the difference between a lock-mint and a burn-mint bridge?
Lock-mint: asset locked on chain A, minted on chain B. Reversible (can unlock). Burn-mint: asset burned on chain A, minted on chain B. Irreversible (if bridge fails, asset is gone). Lock-mint is safer but slower. Burn-mint is faster but riskier. Most modern bridges use hybrid approaches.
▶How do you secure a bridge's validator set?
Validator security = stake-based consensus (validators risking capital), cryptographic proofs (no faking signatures), slashing rules (bad validators lose stake), and key management (validators' private keys are cold storage, rarely used). Multi-sig (m-of-n validators) adds redundancy: 7-of-10 validators must approve a transfer.
▶What role do zero-knowledge proofs play in bridges?
ZK proofs let a bridge verify a transaction happened on chain A without trusting any single validator. Example: chain A produces a ZK proof ("transaction X happened"), bridge verifies the proof (not the transaction), and mints on chain B. Eliminates validator collusion risk. But ZK proof generation is slow (5-30 sec), not suitable for real-time trading.
▶How do you test a bridge for security?
Formal verification (prove code is correct mathematically), fuzzing (send random inputs, look for crashes), penetration testing (try to steal funds with 10+ attacks), and live-net audits (monitor real transfers for anomalies). Never ship a bridge without 2+ external audits.
▶What's a reasonable slashing penalty for validators?
If a validator loses $10M in bridge TVL from bad behavior, penalty should be $5-10M (50-100% of losses). Too low = validators have cheap insurance to steal. Too high = no one will validate. The penalty should be painful enough to deter collusion but not bankrupt an honest validator who makes a mistake.
▶How do you choose between centralized and decentralized bridge architecture?
Centralized (e.g., Coinbase-run bridge): fast, simple, but Coinbase could steal funds. Decentralized (multi-sig validators): slower, complex, harder to steal funds. Most projects start centralized, upgrade to decentralized as TVL grows. For <$100M TVL, centralized is acceptable. For >$1B TVL, must be decentralized.