The data suggests a troubling pattern. Cross-rollup bridge TVL grew 340% in Q3 2025, yet audit coverage for these contracts averaged 60% less than their L1 counterparts. Tracing this anomaly back to the EVM, I found something more disturbing: the majority of these bridges deploy with the same optimistic verification logic that failed on Ronin and Wormhole, just wrapped in a newer UI.
The market is treating interoperability as a solved problem. The code says otherwise.
The Expansion Imperative
Layer-2 networks have evolved from a scalability experiment into a full-blown land grab. OP Stack forks alone account for over 40% of all rollup deployments in 2025. Every major ecosystem—Coinbase, Kraken, Sony—has launched or announced their own chain. The rationale is simple: vertical integration captures more value than horizontal participation.
This creates a structural problem. Each new rollup is a separate settlement domain with its own state root, its own fraud proof window, and its own sequencer. The moment you move assets between them, you're no longer transacting within a single trust domain. You're crossing a bridge.
The industry narrative treats this as a UX inconvenience. Tracing the security implications back to the EVM reveals it as an architectural regression.
Core Mechanics: The Verification Gap
Consider the standard bridge architecture. Lock assets on L1, mint a representation on L2, or burn and mint between two L2s. The verification layer determines how much you trust the operator. Two dominant patterns exist:
Optimistic verification: Assume valid, challenge within N days. Security depends on watcher presence, challenge period length, and the cost of mounting a valid fraud proof. The economic model works only if honest actors can profit from catching fraud.
ZK verification: Prove validity at the source. Mathematically sound, computationally expensive. The bottleneck shifts to proof generation time and verification gas costs.
The market has largely settled on a hybrid: optimistic for token transfers, ZK for state commitments. This creates a subtle vulnerability. The optimistic layer validates the transfer, but the ZK layer validates the state. An attacker who can manipulate the mapping between these two layers bypasses both security models.
I spent six months in 2020 simulating fraud proof vulnerabilities on naive optimistic models. The core insight from that research remains relevant: any delay between state commitment and final verification creates a window for adversarial replay. Cross-rollup bridges widen this window.
The Economic Disincentive
Here's the counter-intuitive part. The economic incentives for securing cross-rollup bridges are fundamentally weaker than for L1 security. On Ethereum, validators stake 32 ETH and face slashing. On a bridge, the "validators" are often multi-sig signers with minimal at stake. The security budget is a one-time deployment cost, not a recurring operational expense.
Tracing the gas cost anomaly back to the EVM, consider what happens during a bridge transaction. You're paying for L1 calldata, L2 execution, bridge operator fees, and finality delays. The bridge operator's profit margin comes from these fees. Their incentive is throughput, not adversarial robustness. Security is a cost center, not a revenue generator.
This is the fundamental misalignment. Bridges generate revenue proportional to volume. Security expenditure is fixed. The rational operator optimizes for volume. The rational attacker targets the operator's weakest security point. These curves intersect at the point of exploitation.
Contrarian Angle: The Sequencer Fallacy
The prevailing narrative suggests that decentralized sequencers will solve the bridge security problem. This is a category error. Decentralizing the sequencer makes the production of blocks more robust, but it does nothing for verification of cross-domain state transitions.
A decentralized sequencer produces blocks faster. It does not make fraud proofs cheaper, challenge periods shorter, or ZK proof generation faster. The security bottleneck in cross-rollup bridges is not block production—it's state verification. The two problems are orthogonal.
The market rewards sequencer decentralization because it's measurable and visible. It ignores verification improvements because they're incremental and complex. This is a classic security theater dynamic.
The Threat Model
The realistic attack vector is not a direct bridge exploit. It's a two-step attack. First, compromise a lower-value rollup with weak governance. Second, use that rollup's bridge position to attack a higher-value destination.
Tracing the economic incentives: an attacker spends $10M to compromise a small rollup's governance. They then use that position to drain $500M from a major bridge. The profit margin justifies the effort.
This vertical attack surface doesn't exist on monolithic chains. It's uniquely a product of the rollup-centric ecosystem.
Forward Outlook
The security model for cross-rollup bridges will eventually consolidate around ZK proofs. The math is unforgiving: optimistic models require an honest majority assumption, which breaks under adversarial economic pressure. ZK proofs require no such assumption. The transition will be slow, painful, and marked by at least one major exploit before the market wakes up.
Based on my audit experience, I'd expect the first major cross-rollup bridge exploit to occur within the next 12-18 months. Not because the technology is fundamentally broken, but because the economic incentives for security are structurally misaligned.
The data suggests we're racing toward this outcome. The only variable is which bridge gets hit first. Architecture reveals the true intent. The intent, currently, is speed over security. That's a trade-off the market will eventually regret.