Audited DeFi protocols lost $885M to attacks that occurred completely outside their audit scopes

Researchers from ack3 and the Czech Technical University examined 135 reported DeFi incidents from January to June 2026, totaling $939.86 million in losses. Of the 68 incidents with identifiable pre‑incident audits, 46 attack paths fell completely outside the audit scope, accounting for $680.97 million of the $721.24 million loss in that subset.

Key findings from the study

• Only 68 of 135 incidents had a publicly documented audit prior to the attack.
• 46 attacks (67.6% of audited incidents) were outside every identified audit scope.
• These outside‑scope attacks represented 94.4% of the losses in the audited‑incident group.
• Excluding two large outliers (Kelp DAO and Drift Protocol) reduces the outside‑scope loss share to 72.1%.

Illustrative August incidents

ICON Network replay exploit

On August 27, a replay bug in ICON’s withdrawal flow allowed an attacker to submit the same signed message 1,492 times, draining 119.866 million ICX and 531,600 bnUSD. The migration contract had undergone an external audit, but the mismatch between the signed value and the uniqueness check was not covered by any audit report.

aelf runtime compromise

In mid‑August, aelf reported that an unauthorized smart contract could inject .NET assemblies into node execution, potentially enabling host command execution and credential access. Although aelf’s core contracts have been audited, the specific runtime path exploited was not linked to any pre‑incident audit.

Implications for audit assurance

The data highlights a gap between audit coverage and the live, evolving system. An audit badge does not guarantee that all components—upgrades, privileged keys, front‑ends, or operational controls—are covered. Users need detailed, versioned assurance records that specify the exact code, contracts, and components reviewed, as well as any changes made after the audit.

Recommendations for users and projects

• Require transparent disclosure of the audited repository, commit hash, deployed addresses, and excluded components.
• Track upgrades, key rotations, and runtime isolation mechanisms after an audit.
• Implement real‑time monitoring and automated shutdown triggers to limit damage when prevention fails.