2014 · United States · I - Accommodation and food service activities · Data theft during a multi-year undetected intrusion (espionage-type)
Analysed 2026-11-07 by Hermes
Open this case in the tool →Data theft during a multi-year undetected intrusion (espionage-type)
Strategy to Execution
Evidence coverage: 5/6 stages, weight 45/60
Decisive stage: detection (clearest evidence) · Alternative: Strategy to Leadership
(1) Strategic expectation: as the world's largest hotel group after the September 2016 Starwood acquisition, Marriott carried explicit, formalised safeguards - privacy-policy promises of reasonable security, internal policies and contractual obligations requiring MFA for remote access to sensitive environments, PCI DSS compliance with Reports on Compliance, a funded monitoring programme (SOC, SIEM, IBM Guardium), and a two-year integration plan to bring Starwood's systems up to Marriott standards by December 2018. (2) In practice: intruders compromised an external-facing Starwood webserver on 28 July 2014, harvested privileged credentials, ran malware on over 480 systems across 58 locations and remained undetected until 7-8 September 2018 - including two years under Marriott's ownership and its security review; entire database tables were exported to .dmp files unlogged, 5.25 million passport numbers were held unencrypted, and 339 million guest records were taken. The ICO's penalty notice found four principal failures - insufficient monitoring of privileged accounts, insufficient database monitoring/logging, missing server hardening, and encryption confined to the PCI scope - and held Marriott negligent; the FTC alleged that the very failures the First Breach forensics had surfaced (firewalls, segmentation, access controls, unsupported software, MFA) recurred in the Second Breach. (3) This is a Strategy to Execution gap rather than mere weak security: the expectations sat at the strategic/doctrinal level - written policies, contractual and regulatory obligations, a decided and procured monitoring programme - and the break was in realising them in architecture and configuration: MFA policy not enforced in scope, alerts limited to payment-card tables, logging configurations that the ICO found rendered the SIEM and SOC ineffective, encryption scope set by audit perimeter rather than data sensitivity. It is not Expertise to Execution (the record shows formal artefacts and bought tooling whose implementation was missing, and Marriott bona fide believed, on PCI ROC assurances accepted by the ICO, that CDE access was MFA-protected), and not Strategy to Leadership (no source documents a leadership decision to defer, underfund or override a warning, and the ICO withdrew its provisional notification findings, so disclosure was not mismanaged). The entry vector (webserver compromise) is not load-bearing for the type; the classification rests on the detection stage, whose origin and break are documented in three independent primary sources (ICO penalty notice, FTC complaint, OPC findings), corroborated by the containment and egress stages. RM22 (Strategy to Execution lens): internal policies and contractual obligations required MFA for remote access to sensitive environments, but the requirement never became an enforced, correctly-scoped control - the attacker logged in with harvested privileged credentials through a gap in MFA coverage and former-employee accounts were not terminated (causal role: Enabling). RM30 (Strategy to Execution lens): segmentation was a standing architectural requirement (PCI CDE separation, Marriott integration standards), yet the architecture did not contain the compromise: intruders moved easily between hotel-property systems and the corporate network, reaching 480+ systems across 58 locations (causal role: Amplifying). RM33 (Strategy to Execution lens): logging requirements that the company had formally decided (SOC, SIEM, Guardium) were not implemented in the systems - no logging of privileged user activity, no log aggregation, no logging of file creation or bulk table exports - so the intrusion ran four years undetected and Marriott could not determine what data was taken (causal role: Obscuring). RM26 (Strategy to Execution lens): the data-protection requirement to encrypt sensitive personal data was applied only within the PCI audit perimeter, leaving 5.25 million passport numbers unencrypted, card data outside the secure environment and a decryption script inside the same database, so the exfiltrated volume was far more harmful than necessary (causal role: Amplifying). A further condition, the recurrence of the identical failure set after the 2014/2015 First Breach despite documented forensics and Marriott's pre-closing review of those failures, is described here rather than coded under a fifth marker.
RM22 — Authentication / credential exposureRM30 — Segmentation / containment weaknessRM33 — Logging / forensic-readiness gapRM26 — Data-protection / cryptographic weaknessLink check: the row's FTC source URL (ftc.gov/news-events/news/press-releases/2024/10/ftc-takes-action-against-marriott-starwood-over-data-breaches) returns 404 and the ICO July 2019 statement-of-intent URL returns 403 (bot wall); replaced with the underlying documents of the same actions - the FTC complaint and Analysis to Aid Public Comment of the October 2024 FTC action (File 1923022) and the ICO's final Penalty Notice of 30 October 2020, opened via a web.archive.org snapshot (fine reduced from the intended GBP 99.2m to GBP 18.4m). ICO MPN scope: the fine addresses the period May-September 2018 under GDPR; the 2014-2018 timeline and failure list come from the FTC complaint and OPC report. The ICO expressly dropped two provisional findings - the MFA coverage gap (Marriott's reliance on PCI Reports on Compliance was reasonable) and the Art. 33/34 notification breaches - so no disclosure-failure marker is coded. The 2024 FTC action is a consent order with injunctive relief; the USD 52m payment was agreed with the 49 state AGs and DC (FTC AAPC; Marriott press reporting). Actor: unattributed in the regulator documents; attribution to Chinese state actors rests on press reports of US government statements (secondary). Related row 984 (informationisbeautiful) describes the same Starwood reservation-system incident per its title; its case file row-984.json is not present in cases-run500 and no result-984.json existed in the output directory, so naming could not be aligned. An alleged 2022 joint government advisory citing a 5.2-billion-record total could not be located or opened and is not cited; record counts follow the FTC/ICO (339 million).