Cross-Chain Bridge Security Models Compared
Compare cross-chain bridge security models: trustless, federated, optimistic, and ZK-proof. See exploit history, trust assumptions, and Bitcoin bridging options.
Bridge Security Models at a Glance
Cross-chain bridges have become critical infrastructure for moving assets between blockchains, but they remain one of the most exploited categories in crypto. Between 2021 and Q3 2025, Global Ledger documented 34 confirmed bridge incidents accounting for nearly $2.9 billion in stolen assets. The root cause in most cases was not a novel cryptographic break but a failure in the bridge's trust model: compromised validator keys, insufficient signer thresholds, or unchecked proof verification.
Bridge designs fall into four broad categories based on how they verify that a transaction on the source chain actually happened before releasing assets on the destination chain. Each model makes different tradeoffs between security, speed, cost, and decentralization.
| Security Model | Verification Method | Trust Assumption | Latency | Capital Efficiency | Major Exploits |
|---|---|---|---|---|---|
| Trustless (Light Client) | On-chain consensus proof verification | Source chain consensus only | Minutes to hours | High | None confirmed |
| Federated (Multisig/MPC) | Committee of validators signs attestations | Honest majority of signers | Seconds to minutes | High | Ronin, Wormhole, Harmony, Multichain |
| Optimistic | Assumed valid unless disputed within window | At least one honest watcher | Hours (dispute window) | Medium (relayer capital) | None confirmed |
| ZK-Proof | Zero-knowledge proof of source chain state | Proof system soundness | Minutes (proof generation) | High | None confirmed |
Trustless Bridges: Light Client Verification
Trustless bridges run a light client of the source chain on the destination chain, verifying block headers and Merkle proofs directly in a smart contract. The receiving chain checks that a transaction was included in a finalized block without relying on any external committee. The security of the bridge reduces to the security of the two chains themselves.
The Inter-Blockchain Communication protocol (IBC), used across the Cosmos ecosystem, is the most deployed example. IBC light clients track the counterparty chain's consensus state and verify every packet proof cryptographically. Relayers transport messages between chains but cannot forge or alter them: a faulty relayer can delay delivery but cannot fabricate transactions. IBC also includes misbehaviour detection that freezes a client if it encounters non-monotonic timestamps or conflicting headers from a compromised counterparty.
The tradeoff is engineering complexity and gas cost. Verifying full consensus proofs on-chain is expensive, particularly on Ethereum where calldata and computation carry significant costs. This has motivated the shift toward ZK-compressed light clients, where a zero-knowledge proof attests to the validity of a batch of headers, reducing on-chain verification to a single proof check.
Federated Bridges: Multisig and MPC
Federated bridges rely on a committee of validators who collectively attest that a deposit on the source chain occurred before minting or releasing assets on the destination chain. This is the most common bridge design and also the most exploited. The security assumption is that a threshold of signers remains honest and uncompromised.
Wormhole uses a Guardian network of 19 named validator operators (including Figment, Everstake, and Google Cloud), requiring 13-of-19 signatures to approve a message. Each Guardian independently observes events on supported chains and produces a Verifiable Action Approval (VAA). The February 2022 exploit bypassed signature verification entirely, allowing the attacker to forge Guardian approvals and drain approximately $326 million.
Multichain used MPC (multi-party computation) to distribute key management across nodes, but the protocol suffered a catastrophic $130 million drain in July 2023 after its CEO disappeared and insider access to key material was suspected. The Harmony Horizon bridge used a minimal 2-of-4 multisig threshold, which collapsed when attackers compromised just two private keys and drained $100 million in June 2022. The FBI later attributed this attack to North Korea's Lazarus Group.
Key takeaway: Federated bridges fail catastrophically when the signer set is too small, too concentrated, or operationally compromised. The difference between a 2-of-4 multisig and a 13-of-19 guardian set is the difference between $100 million lost and a survivable security posture.
Optimistic Bridges: Dispute-Based Verification
Optimistic bridges treat cross-chain messages as valid by default and rely on a challenge period during which any observer can dispute a fraudulent claim. This model borrows from optimistic rollups: security depends on at least one honest watcher monitoring the system and submitting a fraud proof during the dispute window.
Across Protocol is the leading optimistic bridge. Relayers front their own capital on the destination chain to fill user transfers immediately, then submit repayment claims in batches to UMA's optimistic oracle on Ethereum. A data worker posts the batched assertion with a financial bond. If no one disputes the claim within the challenge window (typically 1 to 2 hours), the relayer is reimbursed from the protocol's liquidity pool. If a dispute is raised, the challenger posts a matching bond and resolution escalates to UMA's Data Verification Mechanism (DVM), where UMA token holders vote on the outcome.
Connext (now rebranded as Everclear) used a similar optimistic model with a centralized sequencer that computed aggregate roots off-chain, monitored by watchers who could halt the system if they detected a forged message. Everclear is migrating to an Arbitrum Orbit-based rollup with Hyperlane for cross-chain messaging.
The optimistic model's strength is that it requires only one honest participant to maintain security. Its weakness is latency: the dispute window introduces delays that make the bridge slower than federated alternatives for settlement finality, even though users experience fast fills thanks to relayer capital.
ZK-Proof Bridges: Cryptographic Verification
ZK-proof bridges use zero-knowledge proofs to cryptographically verify source chain state transitions on the destination chain. Rather than trusting a committee or waiting for a dispute window, the destination chain verifies a succinct proof that the source chain's consensus was followed correctly. The trust assumption reduces to the mathematical soundness of the proof system itself.
Polyhedra's zkBridge uses zkSNARKs to prove Ethereum's consensus on other chains. The team developed deVirgo, a distributed prover, and recursive proof composition to reduce both proof generation time and on-chain verification cost. Succinct's SP1 zkVM takes a more general approach: in April 2025, a ZK light client built on SP1 secured IBC transfers from Cosmos to Ethereum, and in July 2025, the Across bridge moved its interoperability stack onto SP1. Succinct's prover network launched on mainnet in August 2025.
ZK bridges address the core vulnerability of federated designs (no trusted committee) and the latency problem of optimistic designs (no dispute window). However, they introduce their own risks: proof system bugs, prover liveness dependencies, and the complexity of correctly encoding consensus rules into circuits. No major ZK bridge has suffered a confirmed exploit to date, but the technology is younger and less battle-tested than alternatives.
Bitcoin Bridging: A Different Security Landscape
Bitcoin lacks native smart contract programmability for on-chain proof verification, which means bridge designs that work on Ethereum (light clients, ZK verifiers) cannot be directly replicated. Bitcoin bridges must use off-chain mechanisms to custody BTC and mint representations on other chains. This creates a spectrum of trust models specific to wrapped Bitcoin.
| Bridge | Model | Custody | Threshold | Supply | Exploit History |
|---|---|---|---|---|---|
| wBTC | Custodial | BitGo (single custodian) | N/A (centralized) | ~$10B+ | No exploit (counterparty risk) |
| tBTC v2 | Threshold signatures | Distributed staker-operated wallets | Threshold signing quorum | Growing | Attempted invalid mint blocked (May 2026) |
| sBTC | Signer network | Decentralized Stacks signers | Signer set consensus | Growing | None confirmed |
| USDB on Spark | Native L2 | No wrapping needed | Spark protocol consensus | Emerging | None |
wBTC remains the most liquid wrapped Bitcoin with over $10 billion in supply, but its security depends entirely on BitGo as a single custodian. BitGo's 2024 restructuring with a Hong Kong-based partner drew criticism from the community over jurisdiction and counterparty risk. tBTC v2 distributes custody across stakers with signing authority over Bitcoin wallets, and in May 2026 its protocol safeguards reportedly blocked an attempted uncollateralized mint. sBTC on Stacks uses a decentralized signer network that is planned to transition to full decentralization over time. For a deeper comparison of wrapped Bitcoin models, see our wrapped Bitcoin security analysis.
An alternative approach avoids wrapping entirely. Spark operates as a Bitcoin Layer 2 that supports stablecoins like USDB natively, without bridging BTC to another chain. This eliminates bridge custody risk for users who want dollar-denominated value on Bitcoin without the trust assumptions of wrapped assets. For more on how Spark compares to other Bitcoin L2 designs, see the Bitcoin Layer 2 bridge comparison.
Major Bridge Exploit Timeline
The following table documents the largest confirmed bridge exploits. Chainalysis estimated that approximately $2 billion was stolen across 13 bridge hacks in 2022 alone, accounting for roughly 69% of all crypto stolen that year.
| Date | Bridge | Loss | Security Model | Root Cause |
|---|---|---|---|---|
| Feb 2022 | Wormhole | ~$326M | Federated (19 Guardians) | Signature verification bypass allowed forged approvals |
| Mar 2022 | Ronin (Axie Infinity) | ~$624M | Federated (5-of-9 multisig) | Spear-phishing compromised 5 validator keys; undetected for 6 days |
| Jun 2022 | Harmony Horizon | ~$100M | Federated (2-of-4 multisig) | Two private keys compromised; attributed to Lazarus Group |
| Aug 2022 | Nomad | ~$190M | Optimistic (Merkle proof) | Routine upgrade initialized trusted Merkle root to zero, allowing arbitrary withdrawals |
| Jul 2023 | Multichain | ~$130M | Federated (MPC) | CEO disappearance; suspected insider access to MPC key material |
| Apr 2026 | KelpDAO adapter | ~$292M | Bridging adapter | Drained rsETH from bridging adapter contract |
A pattern emerges: four of the six largest exploits targeted federated (multisig/MPC) bridges. The Nomad exploit is the exception, where an implementation bug in an optimistic design allowed anyone to drain funds. The KelpDAO incident in 2026 targeted a bridging adapter rather than a core bridge protocol. For detailed postmortems of the 2022 exploits, see our bridge hack postmortem analysis.
How to Evaluate Bridge Security
When choosing a bridge, evaluate these dimensions:
- Signer set size and distribution: a 2-of-4 multisig is fundamentally weaker than a 13-of-19 guardian set with named, independent operators
- Verification method: on-chain proof verification (light client or ZK) eliminates the trusted committee entirely
- Audit history: check whether the bridge has undergone multiple independent audits and whether findings were addressed
- Escape hatches: can users withdraw assets if the bridge operator disappears or the validator set goes offline
- Time in production: newer ZK bridges have strong theoretical properties but less battle-testing than established federated designs
- Bug bounty program: bridges with large, active bounty programs incentivize white-hat disclosure over exploitation
For a broader comparison of bridge fees and routes, see the bridge security comparison tool and the cross-chain swap comparison.
Frequently Asked Questions
Which bridge security model has never been hacked?
As of late 2026, no major trustless (light client) or ZK-proof bridge has suffered a confirmed exploit. IBC's light client design has processed billions in cross-chain transfers across the Cosmos ecosystem without a protocol-level security failure. However, "never been hacked" does not mean "unhackable": ZK bridges are younger and less battle-tested, and implementation bugs remain possible in any software system. The Nomad exploit showed that even a theoretically sound design can fail due to an initialization error.
How does Bitcoin bridging differ from Ethereum bridging?
Ethereum supports arbitrary smart contract logic, which allows bridges to verify consensus proofs, run light clients, and validate ZK proofs directly on-chain. Bitcoin's scripting language is intentionally limited, so bridge security on Bitcoin relies on off-chain custody mechanisms: single custodians (wBTC), threshold signature schemes (tBTC), or signer networks (sBTC). This is why Bitcoin Layer 2 protocols like Spark take a different approach, enabling stablecoin transfers natively without wrapping or bridging BTC to another chain.
What is the safest way to bridge assets cross-chain?
The safest approach depends on the chains involved and the amount being transferred. For EVM-to-EVM transfers, bridges using ZK-proof or light client verification eliminate trusted committees entirely. For large transfers, using the canonical bridge of a rollup (which inherits the security of the underlying L1) is generally preferred over third-party bridges, though it comes with longer withdrawal times. For Bitcoin, avoiding bridging entirely by using native Layer 2 protocols is the lowest-risk option.
Why have federated bridges been exploited so frequently?
Federated bridges concentrate security in a small set of signing keys. If an attacker compromises enough keys to meet the signing threshold, they gain full control over the bridge's locked assets. The Ronin exploit required only 5 of 9 keys. Harmony Horizon required only 2 of 4. Operational security of the signer set becomes the bridge's single point of failure, and bridge exploits have shown that social engineering, insider threats, and poor key management are recurring attack vectors.
What is an optimistic bridge dispute window?
An optimistic bridge treats cross-chain messages as valid unless someone challenges them during a fixed time period called the dispute window. In Across Protocol, this window is typically 1 to 2 hours. During that time, any observer can post a bond and challenge a fraudulent relayer claim, which escalates to UMA's token-holder voting mechanism for resolution. The dispute window adds latency but means the bridge only needs one honest watcher to remain secure.
Are ZK-proof bridges the future of cross-chain security?
ZK-proof bridges offer the strongest theoretical security guarantees because they replace trusted committees with mathematical proofs. Projects like Polyhedra's zkBridge and Succinct's SP1 have demonstrated working ZK light clients in production. However, ZK bridges face practical challenges: proof generation requires specialized hardware, prover liveness is a dependency, and correctly encoding consensus rules into circuits is error-prone. The technology is maturing rapidly, and the trend across the industry is clearly toward ZK-based verification as proving costs decrease.
How much has been lost to bridge exploits in total?
Between 2021 and Q3 2025, Global Ledger documented approximately $2.9 billion in losses across 34 confirmed bridge incidents. Chainalysis estimated that bridge hacks accounted for roughly $2 billion of the total in 2022 alone, representing 69% of all crypto stolen that year. PeckShield reported an additional $328 million in bridge losses through mid-May 2026. These figures come from secondary aggregators and should be verified against primary sources like Chainalysis, Immunefi, or PeckShield for the latest totals.
This tool is for informational purposes only and does not constitute financial or security advice. Exploit data is based on publicly available reports from Chainalysis, PeckShield, Immunefi, Global Ledger, and project postmortems. Loss figures are approximate and may vary by source. Bridge security models evolve over time: always verify the current design and audit status of any bridge before transferring assets.
Build with Spark
Integrate bitcoin, Lightning, and stablecoins into your app with a few lines of code.
Read the docs →
