Worked example · 2015-02

BTER hot-wallet drain — Bitcoin — 2015-02-14

Loss: approximately 7,170 BTC (~$1.65M–$1.75M at the at-time BTC price of ~$235–$245). The drain emptied the BTC hot-wallet float of BTER, a Chinese-incorporated multi-asset altcoin exchange operating out of Shanghai.
Recovery: partial. Approximately 555 BTC of the stolen balance was returned to BTER from intermediary addresses over the months following the event (per BTER's own public statements), and the operator publicly offered a 720-BTC bounty (~10% of the stolen total) for the return of the remainder — an unusual structural choice on the public record at the time. The bulk of the stolen funds was never recovered. BTER absorbed the residual loss against its own balance sheet, suspended BTC trading temporarily, restored customer balances from cold-wallet reserves, and remained operational; the operator subsequently rebranded part of its product surface as Gate.io in 2017.
OAK Techniques observed: OAK-T11 broadly construed (custody-side compromise — operator-internal access to credentials covering cold-wallet handling). BTER's own contemporaneous statement attributed the event to compromise of credentials used in cold-wallet handling — the entry vector therefore sits at the operator-internal end of the T11 family rather than at the third-party-vendor end (OAK-T11.001) or the wallet-software-distribution end (OAK-T11.002). No exact OAK v0.1 sub-technique cleanly fits.
Attribution: pseudonymous — no public-record indictment, attribution-cluster claim, or law-enforcement disposition has named the BTER 2015-02 attacker. The case predates the post-2017 DPRK-dominated South-Korea-targeting cohort and sits cleanly outside the OAK-G01 frame.
Key teaching point: BTER's public 720-BTC bounty for return of the stolen funds is one of the earliest examples on the public record of an exchange operator pricing a return-incentive offer as a standalone recovery instrument distinct from law-enforcement action or on-chain forensics. The pattern — operator-funded bounty, denominated as a fraction of the stolen total, broadcast publicly to the attacker's known on-chain footprint — recurs episodically across the post-2020 DeFi-incident record (Poly Network 2021, Euler 2023, and adjacent cases) but BTER 2015 is the foundational anchor.

Summary

BTER (Bitcoin Trading Exchange Resource) was, in 2015, one of the larger Chinese-jurisdiction multi-asset cryptocurrency exchanges, operating a wide listing surface across Bitcoin and a long tail of altcoins. On 2015-02-14 the exchange's operator publicly disclosed that approximately 7,170 BTC had been drained from its Bitcoin hot wallet through what the operator described as compromise of credentials that controlled cold-wallet-handling operations. Per BTER's own contemporaneous statements, the credentials in question were not raw private keys but a credential-set covering the operational workflow that moved funds between cold-wallet reserves and the BTC hot wallet — meaning the attacker reached the operator's custody-management workflow rather than directly compromising the cold-wallet keys themselves.

BTER suspended BTC deposits and withdrawals on the day of disclosure and announced two recovery actions in parallel: a 720-BTC public bounty (~10% of the stolen total) for return of the funds, broadcast to the attacker's known on-chain footprint, and a request for cooperation from Chinese law-enforcement (Shanghai municipal police) and from major Chinese-jurisdiction exchanges (BTCChina, Huobi, OKCoin) to flag the attacker-controlled addresses if proceeds appeared on those venues. Approximately 555 BTC was eventually returned to BTER from intermediary addresses (per BTER's later statements); the bulk of the loss was not recovered. BTER absorbed the residual loss corporately, restored customer BTC balances from cold-wallet reserves, and resumed BTC trading after a multi-day operational pause. The exchange remained operational through the rest of 2015 and into the post-2015 Chinese regulatory turbulence, eventually rebranding part of its product surface as Gate.io in 2017.

The case is a foundational T11-broadly-construed worked example for OAK on three axes: it predates virtually all of the post-2017 exchange-custody vocabulary; it sits at the operator-internal-credential-compromise end of the T11 family rather than at the third-party-vendor end (where Bitfinex 2016 sits) or the operator-internal-wallet.dat end (where Mt. Gox 2011–2014 sits); and it is the earliest public-record example of an operator-funded return-bounty as a standalone recovery instrument. The pseudonymous-unattributed attribution surface mirrors Bitstamp 2015 and reinforces the OAK convention that the pre-2017 exchange-hack record is overwhelmingly forensic-only with no actor identification.

Timeline (UTC unless noted)

When	Event	OAK ref
Pre-event (2014 → 2015-02)	BTER operates as a Chinese-jurisdiction multi-asset exchange with a hot-wallet / cold-wallet split for BTC; the cold-wallet-handling workflow is mediated by an operator credential-set that controls movement of funds between the two layers	(custody-design surface created)
2015-02 (pre-disclosure window)	Attacker obtains the credential-set covering BTER's cold-wallet-handling workflow; per BTER's later statement the compromise was of the credential surface rather than the cold-wallet keys directly	T11 entry — operator-internal credential compromise
2015-02-14	Approximately 7,170 BTC drained from BTER's BTC hot wallet to attacker-controlled addresses	T11 extraction — hot-wallet drain via cold-wallet-handling credential access
2015-02-14 (same day)	BTER detects the loss, suspends BTC deposits and withdrawals, announces the breach publicly	(operator detection / disclosure)
2015-02-15 → 2015-02-17	BTER announces a 720-BTC public bounty (~10% of the stolen total) for return of the funds; broadcasts the offer to the attacker's known on-chain footprint and to the Chinese exchange ecosystem	Recovery — operator-funded return-bounty (foundational anchor)
2015-02 → 2015-04	BTER engages with Shanghai municipal police; flagging requests to BTCChina, Huobi, OKCoin and adjacent venues; approximately 555 BTC eventually returned from intermediary addresses per BTER's later public statements	(partial recovery — return-mediated)
2015-02 → 2015-03	BTER restores customer BTC balances from cold-wallet reserves and resumes BTC trading after a multi-day operational pause; residual loss absorbed corporately	(operational recovery — corporate-funded)
2015 → 2017	BTER continues to operate through the rest of 2015 and the 2016–2017 Chinese regulatory turbulence; partial rebrand to Gate.io in 2017	(operator continuity)
2017-09	Chinese government bans domestic cryptocurrency exchanges; Gate.io / BTER product surface migrates offshore as part of the broader industry response	(regulatory-context post-event)

What defenders observed and learned

Pre-event: the load-bearing structural property is that the cold-wallet-handling workflow — the operator-internal procedure for moving funds between cold reserves and the hot wallet — was mediated by a credential-set reachable by an attacker who never touched the cold-wallet keys themselves. The defender lesson is that "cold storage" is an operational property, not just a key-management property: a workflow that periodically authorises egress from cold storage into a hot wallet has its own attack surface (the workflow's credential-set), and that surface can be compromised independently of the cold keys. Modern custody-runbook practice treats cold-to-hot rebalancing as a multi-party, multi-channel, time-boxed operation precisely because the BTER failure shape demonstrated that a credential-set covering the rebalancing workflow is itself a single point of failure.
At-event: detection happened during the extraction (operator-side observation of anomalous hot-wallet outflow on the same day), not before it. As with Bitstamp 2015 and Mt. Gox 2011–2014, the post-2017 industry baseline of automated egress-anomaly alerting was not yet operational in 2015; detection came from accounting-side observation rather than signing-side velocity / volume / pattern checks. The structural shape of the BTER detection latency is the same as Bitstamp's and the same as Mt. Gox's; the difference is magnitude and duration, not detection capability.
Post-event (recovery side): the 720-BTC public bounty is the load-bearing innovation in the case. The recovery instrument BTER broadcast — operator-funded, denominated as ~10% of the stolen total, advertised through Chinese-language exchange channels to the attacker's on-chain footprint — is the earliest public-record exchange-side use of a return-bounty as a standalone recovery mechanism. The instrument worked partially: ~555 BTC returned from intermediary addresses (~7.7% of the stolen total) is in the right magnitude for a bounty-mediated partial-return outcome, though BTER did not publicly attest that the returns were attributable to the bounty offer rather than to law-enforcement pressure or downstream-exchange flagging. Defenders writing exchange-recovery runbooks should treat operator-funded return-bounty as a named option rather than reinventing the mechanism per incident.
Post-event (operator continuity): BTER's continuation as an operating exchange through the loss event is itself a structural feature. The operator absorbed a ~7,170-BTC loss against its corporate balance sheet, restored customer balances from cold-wallet reserves, and remained operational — a recovery shape that recurs at Bitstamp 2015 (smaller loss) and Coincheck 2018 (much larger loss), and that contrasts with the operator-collapse shape of Mt. Gox 2014. The Chinese-jurisdiction operating environment in 2015 — pre-2017-regulatory-ban, before the post-2020 South Korean licensing regime, with no local equivalent of FinCEN-style operator-licensing constraint — is part of why the operator-continuity outcome was achievable; contributors writing jurisdiction-of-incorporation analyses should preserve this context.

What this example tells contributors writing future Technique pages

BTER 2015-02 belongs alongside Mt. Gox 2011–2014 and Bitstamp 2015 as a foundational pre-2017 T11-broadly-construed anchor. The three cases together cover the operator-internal end of the T11 family across three structurally distinct entry sub-vectors: copied-wallet.dat private-key file (Mt. Gox), administrator-endpoint compromise via spear-phishing (Bitstamp), and cold-wallet-handling-workflow credential compromise (BTER). Contributors writing T11 sub-technique pages or future T11.x additions should treat the three as a triplet that defines the operator-internal sub-vector range, with the BTER credential-workflow vector as the case that most clearly demonstrates that "cold storage" can be defeated through the workflow that touches it rather than through the keys themselves.
The operator-funded return-bounty pattern is its own recovery-mechanism sub-class and should be documented as such. The pattern — operator broadcasts a denominated fraction of the stolen total as a return-incentive, addressed to the attacker's on-chain footprint — recurs across the post-2020 DeFi-incident record (Poly Network 2021, Euler 2023, and adjacent cases). BTER 2015 is the foundational anchor on the public record. Contributors writing future recovery-mechanism analysis or worked examples in which a return-bounty figures should treat BTER as the chronological anchor and avoid framings that imply the mechanism is a 2020s-era innovation.
pseudonymous-unattributed is the right marker for pre-2017 Chinese-jurisdiction exchange events. No public-record indictment, attribution-cluster claim, or law-enforcement disposition has named the BTER 2015-02 attacker. Contributors writing other pre-2017 Chinese-jurisdiction worked examples should expect a similar attribution surface. The Chinese law-enforcement record on cryptocurrency-incident response from this period is sparse in English-language secondary sources; OAK contributors with access to Chinese-language primary sources are encouraged to expand the attribution-axis material where the record supports it.
The cold-wallet-handling-workflow credential surface deserves an explicit T11.x sub-technique candidate in a future v0.x update. The current OAK v0.1 T11 family does not contain a sub-technique that cleanly captures "compromise of operator credentials covering the workflow that moves funds between cold and hot custody layers, without compromise of the cold-wallet keys themselves." This sub-vector is structurally distinct from third-party-vendor compromise (T11.001), wallet-software distribution (T11.002), and in-use multisig manipulation (T11.003). BTER 2015-02 is the founding case; future v0.x contributors considering a T11.x addition for cold-to-hot workflow credential compromise should treat BTER as the chronological anchor and the present case as the canonical illustration.

Public references

[bterpress2015] — BTER (Bitcoin Trading Exchange Resource). Public statement on the 2015-02-14 hot-wallet incident. 2015-02-14 / 2015-02-15 announcements; primary-source operator disclosure of the breach, the attribution to cold-wallet-handling credential compromise, and the 720-BTC return-bounty offer.
[coindeskbter2015] — CoinDesk. Chinese Bitcoin Exchange BTER Loses 7,170 BTC in Reported Hack. 2015-02-15; contemporaneous English-language press coverage of the disclosure.
[btcmanagerbter2015] — BTC Manager / contemporaneous Chinese-jurisdiction press. BTER hot-wallet incident and 720 BTC bounty offer. 2015-02; secondary-source coverage of the bounty mechanism and the partial recovery posture.
[reutersbter2015] — Reuters. Chinese bitcoin exchange BTER says hackers took 7,170 bitcoins. 2015-02-16; second-source contemporaneous coverage at the moment of disclosure.
[gateioheritage] — Gate.io public-history material covering the 2017 rebrand from BTER, included for operator-continuity cross-reference.

Discussion

BTER 2015-02 sits in a tightly-defined slot in the OAK record: it is the third foundational pre-2017 T11-broadly-construed worked example, after Mt. Gox 2011–2014 and Bitstamp 2015, and it covers a sub-vector of the T11 family — operator-internal cold-wallet-handling-workflow credential compromise — that the other two foundational cases do not cover. The structural failure mode the case demonstrates is operationally distinct from the Mt. Gox failure mode (copied hot-wallet wallet.dat private-key file) and from the Bitstamp failure mode (administrator-endpoint compromise via spear-phishing → lateral movement to signing infrastructure). BTER's failure mode is upstream of the keys: a credential-set covering the workflow that moves funds between cold reserves and the hot wallet was the load-bearing compromised asset, and the workflow's authority to authorise the cold-to-hot rebalancing produced the hot-wallet drain. Contributors writing T11 family analysis should preserve this distinction: the cold-wallet-handling-workflow surface is structurally separate from the cold-wallet-keys surface and from the hot-wallet-signing surface, and a T11.x sub-technique candidate that names the workflow surface explicitly would clarify the family.

The 720-BTC return-bounty is the second load-bearing feature of the case for OAK's purposes. The instrument BTER deployed in 2015-02 is the chronological anchor for a recovery-mechanism pattern that recurs across the post-2020 DeFi-incident record (Poly Network 2021's "Mr. White Hat" return-and-bounty arc, Euler 2023's return-and-bounty arc, and adjacent cases). The 2015 instance differs in form from the post-2020 instances — BTER addressed the offer through Chinese-language exchange channels rather than through on-chain message-field signalling, and the mechanism was operator-broadcast rather than smart-contract-mediated — but the structural shape is the same: operator-funded denominated-fraction return-incentive, broadcast to the attacker's on-chain footprint, in parallel with whatever law-enforcement and downstream-exchange-flagging pressure the operator can mobilise. Contributors writing future recovery-mechanism analysis should treat the BTER instance as the foundational case rather than as a precursor footnote.

The pseudonymous-unattributed attribution surface mirrors Bitstamp 2015's surface and reinforces a point that the post-2018 DPRK-dominated record sometimes obscures: the pre-2017 exchange-hack record is overwhelmingly forensic-only with no actor identification, and the attribution-axis machinery OAK is building must accommodate that surface cleanly. The Chinese-jurisdiction operating environment in 2015 — before the 2017 ban on domestic exchanges, before the post-2020 South Korean licensing regime, and without a local equivalent of FinCEN-style operator-licensing constraint — meant that operator-side incentives to disclose technical detail in primary-source form were even weaker than the contemporaneous global baseline. Contributors writing other pre-2017 Chinese-jurisdiction worked examples should expect a similar pattern: operator disclosure that names the loss magnitude and the broad failure category, but not the technical specifics; English-language press coverage that inherits the operator's framing without independent corroboration; and an attribution surface that remains forensic-only with no public-record actor identification.

Finally, the operator-continuity outcome — BTER absorbing the loss, restoring customer balances from cold-wallet reserves, remaining operational, and eventually rebranding part of its product surface as Gate.io in 2017 — is itself a structurally instructive feature. The recovery shape sits between the corporate-funded reimbursement pattern of Bitstamp 2015 (smaller loss; near-immediate operational recovery) and the operator-collapse pattern of Mt. Gox 2014 (much larger loss; bankruptcy and multi-decade civil rehabilitation). Defenders writing exchange-recovery analysis should preserve the BTER outcome as a documented mid-magnitude reference point and avoid framings that collapse the recovery surface into a binary "exchange survived" / "exchange collapsed" distinction; the operator-continuity outcome at ~7,170-BTC scale is its own case-shape and BTER 2015 is the canonical reference for it.