← Breach analysis by the Dual Vulnerabilities framework

Mozaic Finance insider vault drain (Mozaic developer, 2024)

2024 · Australia · K - Financial and insurance activities · Insider theft of vault funds via stolen private key

Analysed 2026-11-06 by Hermes

Open this case in the tool →

What happened

Insider theft of vault funds via stolen private key

Dual vulnerability type

Expertise to Execution

Evidence coverage: 3/6 stages, weight 21/60

Decisive stage: entry (clearest evidence) · Alternative: Strategy to Execution

Why this classification

(1) Mozaic was a DeFi yield-farming protocol holding roughly $2 million of user funds in its vaults, and its own actions show it held the expectation that privileged vault custody must be protected and vault activity monitored in real time: it had recently upgraded security and was rolling out Hypernative real-time monitoring. (2) In practice an insider - a Mozaic developer - illegally obtained the private keys of a fail-safe security module by compromising a core team member's data, while that legacy module was being phased out but still held live Master-role privileges; through the privileged bridgeViaLifi and execute functions one key emptied both the Hercules and Theseus vaults (~$2.0-2.4M). Detection came from external security firms rather than Mozaic's own tooling, and the contracts were paused only after the funds had drained; recovery and response were otherwise strong (90%+ frozen on MEXC, refund page within hours, full-compensation pledge, criminal charges). (3) The decisive gap is at entry: the remedies that would have prevented it - custody of privileged keys outside ordinary team-member data, key splitting or multisig over the Master role, insider segregation and monitoring - are standard technical/operational specialist knowledge, not strategic commitments, and they were simply not realized in the operation and architecture of the custody system; no source shows leadership deferring, under-resourcing or accepting a known need, so the break is in execution and this is an Expertise to Execution gap. The same evidence can be read as Strategy to Execution (the decided security-upgrade programme not implemented before the breach), and that reading sits as the alternative rather than the decision, because it rests on inference about the planned control while the entry gap is documented by the victim's own detailed disclosure; recovery and response show no gap and are not load-bearing. RM09 (Expertise to Execution lens): specialists could not detect or restrict the insider's activity - a serving developer compromised a colleague's data to exfiltrate custody keys and no insider or endpoint monitoring revealed it (causal role: Enabling). RM20 (Expertise to Execution lens): the security transition itself altered the risk posture beyond what specialists could harden in time - the fail-safe module stayed live with vault privileges while being phased out before the Hypernative measures were operational, and the developer, aware of the upgrades, exploited the limited window that remained (causal role: Enabling). RM04 (Expertise to Execution lens): custody of both vaults concentrated on a single fragile identity - the Master role wallet - a dependency specialists did not reduce, so one key compromise forced the complete drain of Hercules and Theseus at once (causal role: Amplifying). RM34 (Expertise to Execution lens): the available detection tooling was not yet configured over the legacy environment when the drain ran - external firms (Hypernative and others) noticed the exploit first and the pause came after the funds had left (causal role: Obscuring).

Evidence by stage

Risk markers

Sources

Notes

Source caveat: the dataset source_url (an X post) cannot be opened from this container; the victim's full security-incident statement is quoted verbatim on the cited web3isgoinggreat.com archive page, including a screenshot of the original post. Loss figures differ across sources: $2,012,789 (Crypto Briefing), $2.1M (CertiK), $2.4M (Cointelegraph headline); the row name says $2 million and the row description ~$2.4 million. Actor: EuRepoC records 'Not attributed'; the only identification is the victim's own statement that the exploiter was a Mozaic developer, terminated with criminal charges pursued; no public court outcome was found (eurepoc_extra legal_response_*: law-enforcement measures in Australia dated 2024-03-15, consistent with the victim's statement). Country from eurepoc_extra.receiver_country (Australia); press and project documentation describe the project as Sydney-based. The ~12-hour disclosure delay was acknowledged by the victim and attributed to advice not to alarm the actor while internal systems were exposed; it is not coded as RM39 per that marker's exclusion. Description matches the stated victim; no related_rows in the case file. Link check: cointelegraph.com was unreachable at check time (timeout); the page was opened and read during research. Link check: cryptobriefing.com was unreachable at check time (timeout); the page was opened and read during research.