← Breach analysis by the Dual Vulnerabilities framework

Cosmos Bank SWIFT/ATM cash-out (Lazarus Group, 2018)

2018 · India · K - Financial and insurance activities · Theft via payment-system compromise: fraudulent ATM cash-out and SWIFT transfers

Analysed 2026-11-06 by Hermes

Open this case in the tool →

What happened

Theft via payment-system compromise: fraudulent ATM cash-out and SWIFT transfers

Dual vulnerability type

Strategy to Execution

Evidence coverage: 4/6 stages, weight 27/60

Decisive stage: containment (tie-break: pre-breach condition) · Alternative: Expertise to Execution

Why this classification

(1) Strategic expectation: as a SWIFT member running its own ATM card switching, Cosmos Bank stood under SWIFT's mandatory Customer Security Controls Framework, in force for all users since April 2017 and fully effective from January 2018, whose core objectives are to segregate the Swift and critical payment environment from the general IT environment and to detect anomalous activity in transaction records; RBI's IT and cyber-security expectations for banks (the bank itself invoked a July 2018 RBI IT robustness inspection) pointed the same way. (2) In practice: attackers with a foothold in the general network moved laterally to the ATM payment switch server and stood up a malicious proxy switch that broke the link to the core banking system and self-approved 14,849 withdrawals (INR 805 million) across 28 countries; the bank's own core banking system raised no alert and Visa and SWIFT delivered the first warnings; two days after the bank's intervention on Aug 11, the still-unremoved attackers used the bank's own SWIFT Alliance Access credentials to send three fraudulent MT103s (about USD 2 million) to Hong Kong. (3) This is a Strategy to Execution gap rather than merely weak security: a formal, mandatory strategic artefact (environment segregation, secure zone, transaction monitoring) existed for the victim's class before the breach, and the documented reality - payment switch and SWIFT environment reachable and compromised from the general network, no internal alert through the whole first wave - shows it was never realised in the bank's operational architecture; no source shows leadership knew of and deferred a specific control, so the break sits in execution, not leadership. The classification rests on the containment stage as a pre-breach structural condition; the entry vector is unconfirmed and not load-bearing, and the campaign-wide finding that FASTCash switch servers ran unsupported AIX is treated as pattern evidence only. The response stage (attackers not eradicated after detection) is the competing post-breach Expertise to Execution reading and lost the tie-break to the pre-breach condition. As a secondary S2L pattern, the chairman publicly stated the bank's security systems had not been compromised while the bank's own police complaint and press statement described the proxy switch and the misuse of its SWIFT environment. (4) RM30 (Strategy to Execution lens): the mandatory isolation of the payment and Swift environment was not implemented in the bank's architecture - a foothold in the general network reached the payment switch server and the SWIFT Alliance Access environment and let a shadow switch replace the core banking system, so the blast radius covered both theft routes (causal role: Amplifying). RM34 (Strategy to Execution lens): the required transaction-monitoring and alert-handling capability did not operate as intended - the core banking system issued no alert at all, the first warnings came from Visa and SWIFT, and an earlier failed transfer attempt went unnoticed (causal role: Obscuring). RM38 (Strategy to Execution lens): containment and eradication were not executed completely after the Aug 11 detection - persistence and access survived the bank's intervention, so the same intrusion produced the Aug 13 SWIFT theft (causal role: Amplifying).

Evidence by stage

Risk markers

Sources

Notes

Case-file description matches the victim (no data-quality issue). Year 2018 from eurepoc_extra.start_date (10.08.2018). Entry vector unconfirmed: the unsupported-AIX finding in the CISA FASTCash alert is campaign-wide, not incident-confirmed for Cosmos, and is scored as pattern, contributing; expert claims of an outdated ATM switch are unverified allegation (CIO inc.). Detection accounts conflict: the bank said it caught the withdrawals in progress, while police sources said the core banking system gave no alert and Visa/SWIFT alerted first; the police account was used as the more specific. The PwC India forensic report went to the bank and was never published, so the mechanism rests on secondary investigation reporting and vendor analysis; a UNSC panel report (Mar 2019) corroborated the bypass of the three INTERPOL defence layers per press. Chairman's post-incident claim that security systems were not compromised contradicts the bank's own press statement - a public-narrative problem described in the rationale, not coded as RM44 because RM44 requires a pre-breach statement. EuRepoC lists disclosure as by an IT-security company; the bank's own press conference followed on Aug 14. Reuters itself returns 403; the bank's press statement was read in the Reuters reprint at Business Insurance. No related_rows in the case file. Link check: swift.com blocks automated access (HTTP 403); the two cited swift.com pages were opened and read during research via text extraction. The Tata Communications and Forbes leads from the case file are dead or blocked to the link checker and were dropped from the cited sources; nothing in the analysis rests solely on them.