Worked example · 2025-09

UXLINK multisig delegate-call hijack — Ethereum / Arbitrum — 2025-09-22

Loss: approximately $44M extracted from UXLINK's multisig wallet across multiple chains in an attack window beginning on 2025-09-22 and extending into 2025-09-23. UXLINK is a social Web3 / SocialFi project; the drained surface was the project's principal multi-signature wallet contract, with downstream impact across additional ecosystem contracts that the multisig controlled. The headline figure varies across initial reporting (~$11.3M Cyvers initial estimate, ~$30M Hacken estimate) before converging on the ~$44M cumulative figure once the multi-stage extraction pattern (initial admin-control hijack + subsequent token-mint dump) was fully characterised.
Recovery: partial via attacker-side phishing accident. In an unusual follow-on event documented by Scam Sniffer and corroborating on-chain investigators, the UXLINK attacker themselves fell victim to an Inferno Drainer phishing operation on 2025-09-23, losing approximately 542 million freshly-minted UXLINK tokens (~$43M-$50M-equivalent at the time, depending on the price-trajectory window) directly from the attacker's wallet to the Inferno Drainer cluster. The attacker subsequently converted approximately 5,496 ETH to ~11M DAI in March 2026 from the residual proceeds. UXLINK's protocol-level recovery via the operator-side response was minimal; the attacker-side phishing accident is the load-bearing recovery dimension and is structurally unusual in the OAK 2024-2025 corpus.
OAK Techniques observed: OAK-T9.004 (Access-Control Misconfiguration) — primary, in the multisig-delegate-call sub-class: the UXLINK multisig setup permitted arbitrary code execution via a delegate-call vulnerability that bypassed intended signer controls. The multisig had been configured with a 2/x threshold but lacked basic safeguards such as guardians, timelocks, or any kind of governance review process; the attacker exploited the delegate-call path to reassign themselves as owners with a threshold of 1, effectively converting the multisig into a single-signer wallet under attacker control. OAK-T11.002 (Wallet-Software / Operator-Key Compromise — broadly construed) is the entry-vector precondition: per the public reporting, the underlying compromise vector for the initial admin-control surface is incompletely disclosed in the public record, with the attribution centred on the multisig configuration and delegate-call surface rather than on the proximate compromise vector. OAK-T5.003 (Hidden-Mint Dilution — once the attacker converted the multisig into a single-signer wallet under attacker control, they minted approximately 542M UXLINK tokens from nothing and dumped them against on-chain liquidity; the mint was a counterfeit-issuance flow against the protocol's claimed token supply, executed through legitimately-acquired admin privileges).
Attribution: pseudonymous. No public named-individual attribution. The on-chain attacker address was not associated by industry forensic providers with a public OAK-G01 / state-actor cluster as of v0.1. The Inferno Drainer follow-on is the single most distinctive attribution-relevant signal: state-actor / OAK-G01 attackers do not typically lose funds to drain-as-a-service phishing infrastructure, and the loss-of-funds-via-Inferno-Drainer outcome is consistent with a pseudonymous-individual or pseudonymous-small-team attacker profile that lacks the operational-security maturity characteristic of state-actor operations.
Key teaching point: multisig-delegate-call hijack is a recurring 2024-2025 access-control sub-pattern that requires explicit allowlist-constrained call-data validation, not only signer-threshold gating. The defender intuition that "this protocol is secured by a 2/x or M-of-N multisig" is wrong if the multisig contract permits arbitrary delegate-call execution that bypasses the threshold gating. UXLINK is the canonical 2025 worked example for the multisig-delegate-call sub-pattern; the Radiant Capital October 2024 case at /examples/2024-10-radiant-capital.md and the Bybit February 2025 case at /examples/2025-02-bybit.md are the OAK-G01 cohort companions where the attack-payload was modified between the signers' wallet UI / Tenderly simulation surface and the on-chain broadcast surface; the UXLINK case is the structurally-distinct contract-level multisig-delegate-call hijack rather than the signing-time-MITM pattern. The two sub-patterns share the high-level lesson that multisig threshold gating is not a sufficient mitigation primitive on its own; the gating must extend to the call-data surface that the multisig executes against.

Summary

UXLINK is a SocialFi / Web3 social-graph project that issued the UXLINK ERC-20 token across Ethereum and Arbitrum. The project's principal operator-side multisig wallet contract held administrative authority over ecosystem contracts including the UXLINK token's mint authority on Arbitrum. The multisig was configured with a 2/x signer threshold but lacked guardians, timelocks, and governance-review safeguards; the multisig contract code permitted delegate-call execution on caller-supplied target / call-data pairs without sufficient constraint.

On 2025-09-22, an attacker exploited the multisig's delegate-call surface to execute arbitrary code in the multisig's storage / authority context. The attack effectively reassigned the multisig's ownership configuration: the attacker became the sole owner with a threshold of 1, converting the multi-party wallet into a single-signer wallet under attacker control. With administrative authority secured, the attacker drained approximately $11.3M of existing assets from the multisig's ecosystem contracts in the initial phase.

Subsequently, the attacker exercised the now-controlled mint authority on the UXLINK token on Arbitrum to mint approximately 2 billion UXLINK tokens initially, with later estimates suggesting the cumulative mint reached close to 10 trillion UXLINK across multiple subsequent transactions. The attacker dumped the freshly-minted supply against on-chain liquidity, driving UXLINK's spot price down approximately 90% from $0.33 to $0.033. The realised dump proceeds bring the cumulative attacker-side extraction to approximately $44M.

On 2025-09-23, in an unusual reverse-direction event, the attacker fell victim to an Inferno Drainer phishing operation that drained approximately 542 million UXLINK tokens (~$43M-$50M-equivalent at the time of the drain) directly from the attacker's own wallet. The Inferno Drainer cluster — a well-known drain-as-a-service infrastructure documented in OAK at /examples/2024-10-inferno-drainer-handover.md — captured the freshly-minted UXLINK supply that the attacker had not yet dumped. The remaining ETH-denominated proceeds in the attacker's wallet (approximately 5,496 ETH at the time of the September 2025 events) were subsequently converted to ~11M DAI in March 2026.

For OAK's purposes the case is a clean T9.004 multisig-delegate-call-hijack extraction with the unusual structural feature that the attacker themselves became a victim of a downstream phishing operation. The novel OAK contributions are: (a) documenting the multisig-delegate-call sub-pattern as distinct from both function-level access-control bugs (Poly Network 2021) and signing-time-MITM patterns (Radiant Capital 2024, Bybit 2025); (b) capturing the attacker-side-phishing-accident recovery dimension as a structurally unusual recovery class; and (c) identifying the SocialFi / social-Web3 protocol category as a 2025 OAK target surface with thin defender / audit coverage.

Timeline (UTC)

When	Event	OAK ref
Pre-event	UXLINK deploys multisig wallet contract on Ethereum / Arbitrum with 2/x signer threshold; multisig contract permits delegate-call execution on caller-supplied target / call-data without sufficient constraint; no guardians, timelocks, or governance-review safeguards	(T9.004 surface — multisig-delegate-call latent)
Pre-event	Attacker analyses UXLINK multisig contract; identifies the delegate-call surface that bypasses signer-threshold gating	(off-OAK pre-event observation)
2025-09-22 (attack window)	Attacker submits transaction sequence exploiting multisig delegate-call to execute arbitrary code in the multisig's authority context; reassigns ownership configuration to attacker-as-sole-owner with threshold 1	T9.004 extraction (admin-control hijack)
2025-09-22	Attacker drains ~$11.3M of existing assets from multisig-controlled ecosystem contracts	T9.004 extraction (initial)
2025-09-22 / 2025-09-23	Attacker exercises now-controlled mint authority on UXLINK token on Arbitrum; mints ~2 billion UXLINK tokens initially, with subsequent transactions reaching cumulative ~10 trillion UXLINK	T9.004 extraction (mint-authority abuse)
2025-09-22 / 2025-09-23	Attacker dumps freshly-minted UXLINK against on-chain liquidity; UXLINK spot price falls ~90% from $0.33 to $0.033; cumulative dump-realised proceeds reach ~$44M	(downstream dump against AMM liquidity)
2025-09-22 / 2025-09-23	Industry detection chain: PeckShield, Cyvers Alerts, Hacken, SlowMist, Scam Sniffer publish on-chain analyses of the multisig delegate-call hijack and the cumulative-loss figures	(transparency / multi-investigator coverage)
2025-09-23	Attacker falls victim to Inferno Drainer phishing operation; ~~542 million UXLINK (~~$43M-$50M equivalent) drained from attacker's wallet by Inferno Drainer cluster	(attacker-side-phishing-accident; structurally unusual)
2025-09 onward	UXLINK operator team coordinates user communication regarding affected wallets; considers token-restart proposals to compensate users affected by the price collapse	(operator response)
2026-03	Attacker converts residual 5,496 ETH (~$11M+) from the September 2025 events to ~11M DAI; on-chain tracking via standard cluster-correlation tools	(continued laundering pattern)
Continuing	Pseudonymous attribution; no civil-forfeiture action; no named-individual claim; no OAK-G01 cluster correlation as of v0.1	(recovery state)

What defenders observed

The multisig delegate-call surface bypassed the signer-threshold gating that the multisig's documented intent specified. The UXLINK multisig was configured with a 2/x signer threshold — which would, in principle, require two signers to approve any operation against the multisig's authority. The attack exploited a delegate-call execution path that operated outside the threshold gating: a caller-supplied target / call-data pair was executed in the multisig's storage / authority context without requiring the threshold's approval. The defender lesson is that multisig threshold gating is not a sufficient mitigation primitive on its own; the gating must extend to the call-data surface that the multisig executes against, with explicit allowlist-constrained call-data validation that prevents arbitrary delegate-call execution.
The case is structurally distinct from the OAK-G01 signing-time-MITM cohort. The Radiant Capital October 2024 case at /examples/2024-10-radiant-capital.md and the Bybit February 2025 case at /examples/2025-02-bybit.md are the OAK-G01 cohort companions where the attack-payload was modified between the signers' wallet UI / Tenderly simulation surface and the on-chain broadcast surface — the signers approved benign transactions and malicious transactions were broadcast. The UXLINK case is structurally distinct: the multisig contract code itself contained the delegate-call surface that bypassed threshold gating, and the exploit operated at the contract-level rather than at the signing-time level. The two sub-patterns share the high-level lesson (multisig gating is insufficient on its own) but require operationally distinct mitigation surfaces.
The detection chain ran across multiple investigators with substantial estimate divergence. The initial PeckShield / Cyvers Alerts / Hacken estimates ranged from ~$11M to ~$30M depending on which extraction phase (initial drain vs. cumulative drain + mint-and-dump) the estimate covered. The convergence on the ~$44M cumulative figure required several days of cross-cohort analysis. Defender / operator-response practice should expect this estimate-divergence dynamic for multi-stage extraction cases (admin-control hijack + token-mint dump) and should communicate uncertainty bounds explicitly.
The attacker-side-phishing-accident recovery dimension is structurally unusual. The Inferno Drainer follow-on at 2025-09-23 captured the bulk of the freshly-minted UXLINK supply that the attacker had not yet dumped. This is one of the more unusual recovery-state outcomes in the 2024-2025 OAK corpus: the attacker themselves became a downstream victim of a drain-as-a-service phishing operation. Defender / Mitigations-layer practice should treat this dimension as not generalisable — the typical pseudonymous-attacker case does not produce attacker-side-phishing-accident outcomes — but should preserve the dimension as a real (if rare) recovery class.
The SocialFi / social-Web3 category is a 2025 OAK target surface with thin defender coverage. UXLINK is not a DeFi-native protocol in the lending / DEX / yield sense; it is a SocialFi project with an associated token economy. The defender / audit cohort for SocialFi protocols is materially thinner than for DeFi-native protocols, and the operational-security maturity of SocialFi multisig configurations tends to lag DeFi-native protocols by one or more cycles. The UXLINK case is the canonical 2025 evidence; future SocialFi-category incidents should reference UXLINK as the cohort-anchor 2025 case.

What this example tells contributors writing future Technique pages

OAK-T9.004 covers multisig-delegate-call hijack as a distinct sub-pattern. UXLINK September 2025 is the canonical 2025 worked example for the multisig-delegate-call sub-pattern of T9.004. The Prisma Finance March 2024 case at /examples/2024-03-prisma-finance.md is the structurally-related Zap-helper-contract delegate-call case (different context, same underlying delegate-call-with-insufficient-validation pattern). The Li.Fi July 2024 case at /examples/2024-07-li-finance.md is the diamond-pattern facet sub-class companion. The four T9.004 sub-patterns visible in the 2024-2025 corpus are now: (a) function-level access-control modifier missing (Poly Network 2021), (b) diamond-pattern facet (Li.Fi 2024), (c) delegatecall-pattern Zap helper (Prisma 2024), and (d) multisig-delegate-call hijack (UXLINK 2025).
Multisig configuration maturity is a 2025 cohort-level monitoring target. The UXLINK multisig had a 2/x threshold but lacked guardians, timelocks, and governance-review safeguards, and the multisig contract itself permitted arbitrary delegate-call execution. Defender / Mitigations-layer practice should treat multisig configuration maturity as a discrete cohort-monitoring target, with deployment-time verification of (a) signer-threshold gating, (b) timelock / guardian / governance-review safeguards, (c) call-data allowlist constraints on delegate-call surfaces, and (d) per-chain key segregation. The UXLINK case is the canonical 2025 evidence that the multisig-configuration-maturity discipline was not yet operational at coverage sufficient to prevent SocialFi-category incidents.
pseudonymous attribution with attacker-side-phishing-accident recovery is a hybrid recovery status that should be tracked explicitly. Contributors writing recovery-section content should distinguish: (a) on-chain protocol-coordinated recovery (Cetus 2025), (b) attacker-return-of-funds via bounty negotiation (KiloEx, ZKsync, GMX V1, Tapioca, Thala, Munchables), (c) operator-team reimbursement from reserves (DeltaPrime, Banana Gun), (d) validator-coordinated transaction-censorship freeze (Sui / Aptos), (e) validator-coordinated rollback / hard fork (Balancer V2 / Gnosis Chain), and (f) attacker-side-phishing-accident recovery (UXLINK 2025). The sixth class is structurally unusual and not generalisable but should be preserved as a real recovery-state outcome.
SocialFi / social-Web3 protocols are a 2025 target surface with thin defender coverage. Contributors writing OAK target-surface taxonomies should now include SocialFi as a discrete category alongside DeFi-native (lending / DEX / yield), trading-bot (Banana Gun), L1 / L2 infrastructure, and bridge / cross-chain transport. The UXLINK case is the canonical 2025 worked example.
Attribution status pseudonymous should not be conflated with circumstantial-OAK-G01-suspicion. No public technical indicator links the UXLINK case to an OAK-G01 cluster; the attacker's loss-of-funds-via-Inferno-Drainer is itself evidence against state-actor classification (state-actor operations do not typically fall to drain-as-a-service phishing). Contributors writing attribution-section content should preserve the distinction explicitly.

Public references

[uxlinkpostmortem2025] — UXLINK operator-side public statements covering the September 2025 multisig hijack, the affected ecosystem contracts, and subsequent restart-plan considerations.
[theblockuxlink2025] — The Block reporting on the UXLINK multisig exploit including the admin-rights-seizure framing and the token-mint impact.
[bitgetuxlink2025] — Bitget News analysis of the multisig and smart-contract centralisation risks exposed by the UXLINK exploit.
[blockthreatuxlink2025] — BlockThreat (Peter Kacherginsky) Week 39 2025 incident roundup including the UXLINK exploit technical analysis.
[incrypthosuxlink2025] — In-Crypthos technical analysis of the multisig delegate-call vulnerability.
[scamsnifferuxlinkinferno2025] — Scam Sniffer reporting on the UXLINK attacker's subsequent loss to Inferno Drainer phishing on 2025-09-23.
[chainrumorsuxlink2025] — Chain Rumors reporting on the attacker shuffling stolen assets and the ~$43M Inferno Drainer phishing loss.
[panewslabuxlinkinferno2025] — PANews coverage of the 542 million UXLINK token loss to the Inferno Drainer phishing group.
[cryptotimesuxlink2026] — Crypto Times reporting (March 2026) on the UXLINK hacker converting 5,496 ETH to ~11M DAI from the residual 2025 proceeds.
[peckshielduxlink2025] — PeckShield September 2025 monthly hack roundup including UXLINK as the largest single September 2025 incident.

Citations

[uxlinkpostmortem2025] — UXLINK operator-side public statements; primary source for operator timeline and restart-plan considerations.
[theblockuxlink2025] — The Block reporting; primary source for the admin-rights-seizure framing.
[bitgetuxlink2025] — Bitget News; secondary source for multisig-centralisation-risk framing.
[blockthreatuxlink2025] — BlockThreat incident roundup; primary technical source for the multisig delegate-call vulnerability.
[incrypthosuxlink2025] — In-Crypthos; primary technical source for the delegate-call-bypasses-threshold sub-pattern.
[scamsnifferuxlinkinferno2025] — Scam Sniffer; primary source for the Inferno Drainer follow-on event.
[chainrumorsuxlink2025] — Chain Rumors; secondary source for the asset-shuffling and Inferno Drainer phishing-loss figures.
[panewslabuxlinkinferno2025] — PANews; secondary source for the 542 million UXLINK token loss figure.
[cryptotimesuxlink2026] — Crypto Times; primary source for the March 2026 5,496 ETH-to-DAI conversion.
[peckshielduxlink2025] — PeckShield monthly roundup; primary source for the cumulative-loss aggregation and September 2025 cohort framing.

Discussion

UXLINK September 2025 is OAK's canonical case for the multisig-delegate-call hijack sub-pattern of T9.004 and for the SocialFi / social-Web3 target-surface category. The case is the largest single September 2025 DeFi-adjacent incident at ~$44M and is the H2 2025 worked-example anchor for the SocialFi category, which has been thinly represented in the OAK corpus through 2024.

The structural lesson sits at three levels. First, at the contract-level: multisig threshold gating is not a sufficient mitigation primitive on its own; the gating must extend to the call-data surface that the multisig executes against, with explicit allowlist-constrained call-data validation that prevents arbitrary delegate-call execution. Second, at the configuration-level: a 2/x signer threshold without guardians, timelocks, or governance-review safeguards is a configuration that has been repeatedly demonstrated as insufficient across the 2024-2025 corpus, and the cohort-monitoring discipline for multisig configuration maturity is not yet operational at coverage sufficient to prevent SocialFi-category incidents. Third, at the category-level: SocialFi / social-Web3 protocols are a 2025 target surface with thin defender / audit coverage relative to DeFi-native protocols, and the operational-security maturity of SocialFi multisig configurations tends to lag DeFi-native protocols by one or more cycles.

The recovery dimension is the analytically distinctive feature of the case. The attacker-side-phishing-accident at 2025-09-23, where the UXLINK attacker lost approximately 542 million freshly-minted UXLINK tokens (~$43M-$50M equivalent at the time) to the Inferno Drainer cluster, is a structurally unusual outcome that does not generalise across the OAK corpus. The Inferno Drainer cluster is a well-documented drain-as-a-service infrastructure (see /examples/2024-10-inferno-drainer-handover.md); the UXLINK attacker's vulnerability to standard phishing infrastructure is consistent with a pseudonymous-individual or pseudonymous-small-team attacker profile that lacks the operational-security maturity characteristic of state-actor / OAK-G01 operations. Contributors writing future attribution-section content should treat the Inferno-Drainer-follow-on as a strong behavioural signal against OAK-G01 attribution, alongside the absence of forensic-provider-led wallet-cluster correlation.

The attribution status is pseudonymous and should remain so absent new public evidence. Sector-targeting alone (a Web3-adjacent project of substantial nominal valuation) is insufficient grounds for OAK-G01 cluster assignment. Contributors should not over-claim attribution simply because UXLINK is a high-profile target.

For OAK's broader credibility, including UXLINK in v0.1 closes three structural gaps: it adds the September 2025 H2 worked-example anchor to the corpus, it documents the multisig-delegate-call hijack sub-pattern of T9.004 with a 2025 case, and it establishes the SocialFi / social-Web3 protocol category as a discrete target surface with its own operational-security-maturity characteristics distinct from the DeFi-native cohort.