Cross-Chain Bridges: Architecture and Risks

Binance BNB Bridge Oct 2022: $570M lost via forged Merkle proof. Wormhole Feb: $320M via signature bypass. Ronin: $625M. Not coincidences. Bridges most attacked because they aggregate liquidity with complex verification.

Why Bridges Break

Finality and Reorg: Ethereum ~2 epochs (~12 min) for finality. If bridge mints after 1-2 blocks — 3+ block reorg lets attacker get tokens on target with reverted source.

Signature Verification: Most use multisig: N of M validators sign. Wormhole 13/19 guardians. Attack: bug in signature verification itself, not keys.

Lock-and-Mint vs Burn-and-Mint: Lock: original locked on source, wrapped minted on target. Source contract is honeypot with all TVL. One unlock bug = total loss. Burn-and-Mint (Circle CCIP) safer — no locked pool.

LayerZero

LayerZero — arbitrary message passing. Infrastructure for building bridges.

Architecture: Endpoint per chain, Executor (delivers), DVN (Decentralized Verifier Network).

Developer chooses DVN: official or custom. Can require multiple DVN for verification.

OApp base contract for integration. OFT (Omnichain Fungible Token) for tokens.

Wormhole

19 guardians, 13 of 19 threshold. VAA (Verified Action Approval).

Main difference: native non-EVM support (Solana, Aptos, Sui, Algorand).

After 2022: Native Token Transfers (NTT) without locked pool.

Relay and ZK

Relay-based: light client on target verifies source chain. ZK-bridges: ZK-proof of consensus — eliminates committee trust but expensive to verify.

Mandatory Components

Pauser, Rate Limiting, Finality Checks, Relayer Monitoring.

Timelines

Simple bridge: 4–8 weeks. Custom: 12–24 weeks. ZK: 6+ months.