Research

705 exploit cases totaling $78.81B in losses follow a power law (α ≈ 1.33). A handful of catastrophic incidents dominate cumulative damage. Logic Bugs lead with 231 cases, followed by Key Compromise (154) and Reentrancy (84), with attack sophistication escalating from single-tx exploits to orchestrated multi-chain campaigns.

Of 601 intervention-eligible cases ($9.60B at risk), only $2.51B (26.0%) was saved, leaving a $7.09B opportunity gap. Signer Sets handle 71.2% of interventions by count with a median 30-minute response; Delegated Bodies protect $1.10B across 17.3% of cases; Governance votes achieve 73.2% success but take 30+ days.

A "golden hour" separates success from failure: interventions within 60 minutes prevent 82.5% of losses on average; after 24 hours effectiveness drops to 10.9%. The best-performing configuration is Account × Delegated Body. It achieves a 92% containment success rate, while the response-time distribution is starkly bimodal: protocols react in under an hour or not at all.

A stochastic cost model formalizes the intervention decision: Total Cost = Centralization Cost + Blast Rate + Damage Rate × Time. Three empirically validated predictions emerge: tiered authorities outperform pure designs, the optimal response window is under 60 minutes, and scope-limited actions beat protocol-wide pauses. The "Optimistic Freeze" proposal synthesizes these findings into a deployable mechanism.