← Breach analysis by the Dual Vulnerabilities framework

British Airways payment card skimming (Magecart, 2018)

2018 · United Kingdom · H - Transport and storage · Payment card data theft via website skimming after third-party remote-access intrusion

Analysed 2026-11-07 by Hermes

Open this case in the tool →

What happened

Payment card data theft via website skimming after third-party remote-access intrusion

Dual vulnerability type

Strategy to Execution

Evidence coverage: 4/6 stages, weight 35/60

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

Why this classification

(1) British Airways ran one of Europe's largest customer-facing payment estates, bound by PCI DSS commitments and, from 25 May 2018, by GDPR Articles 5(1)(f) and 32; its own Network Access Control Policy of 7 October 2017 mandated multi-factor authentication for all remote network access, explicitly including third-party access, and PCI DSS required file-integrity monitoring and minimisation of stored card data. (2) In practice none of these artefacts had become operational reality: a compromised credential of a Swissport (third-party cargo-services) employee logged into the Citrix Access Gateway without MFA on 22 June 2018 - 13 of the 243 CAG-hosted applications lacked MFA and the risk assessment BA claimed justified the departure could not be located; the attacker broke out of the Citrix environment into network areas never intended for third-party users, read a plaintext domain-administrator credential file, escalated via an enabled Guest account, and moved undetected for roughly ten weeks; a test logging function left active in production had been writing card details including CVVs in plaintext since December 2015; and an injected script copied live payment traffic to the attacker's domain for 15 days until an external party alerted BA on 5 September 2018, after which BA contained the skimmer within 90 minutes and notified the ICO and customers the next day (the ICO proposed a GBP 183.39m penalty and fined GBP 20m in October 2020, finding BA negligent in maintaining the vulnerable systems). (3) This is a knowing-doing gap inside the doctrinal column's diagonal: the strategic artefacts existed - the MFA policy, change management, PCI DSS commitments - but did not translate into architecture and operations; the ICO itself concludes that BA recognised MFA as appropriate and yet failed to implement MFA as required by its own policies, and that each step of the attack could have been prevented or mitigated by available measures, the deficiencies lying in how BA operated its own network rather than in the third party's systems. The classification rests on the entry stage (pre-breach structural condition, winning the tie-break over the containment gap); the route by which the Swissport credentials were themselves compromised was never established but is attacker-side and not load-bearing. (4) RM22 (Strategy to Execution lens): BA's authentication policy mandated MFA on all remote and third-party access, but on the CAG path used in the attack the policy never became an enforced control - 13 of 243 applications were reachable with a single username and password (causal role: Enabling). RM41 (Strategy to Execution lens): least-privilege and privileged-access requirements were not implemented - domain-administrator credentials sat in a plaintext file open to any domain user and a Guest account could be raised to local administrator without detection, turning the initial third-party foothold into domain-wide reach (causal role: Amplifying). RM34 (Strategy to Execution lens): the monitoring requirements BA had undertaken (PCI DSS file-integrity monitoring, logging as the NCSC baseline) were not implemented or operated, so failed administrator logins, the guest-account elevation and the website code changes raised no alert and a third party made the discovery after 103 days (causal role: Obscuring). RM16 (Strategy to Execution lens): PCI DSS 3.1 minimisation and retention obligations were never operationalised - a test logging feature ran live for two and a half years storing CVVs in plaintext, so the attacker found about 108,000 stored cards in addition to the live-skimmed data (causal role: Amplifying).

Evidence by stage

Risk markers

Sources

Notes

Source access caveats: the original ico.org.uk penalty-notice URL returns a Cloudflare challenge to this container, and the live riskiq.com article no longer serves; both are therefore cited as the exact Wayback snapshots opened during this analysis (ICO MPN text-version snapshot 20210415222005; RiskIQ snapshot 20180911154733). The 2020 Wayback copy of the MPN is a scan without a text layer; the cited 2021 snapshot has text. Figures: the row description's 380,000 victims matches RiskIQ's title and BA's initial September 2018 estimate; the ICO's final figure is 429,612 data subjects (244,000 with name, address, card and CVV; 77,000 card and CVV; 108,000 card only from the plaintext logs; plus employee/administrator credentials and up to 612 BA Executive Club accounts). Actor: RiskIQ attributed the skimming to Magecart weeks after the similar Ticketmaster UK incident; the ICO never identified the attackers and BA could not determine how the Swissport credentials were compromised - both attacker-side and not load-bearing. Penalty: the ICO proposed GBP 183.39m in July 2019 and imposed GBP 20m in October 2020, citing mitigation, cooperation and the Covid-19 impact on the airline. Related rows 993 and 1098 (same victim) have no case files in cases-run500 and no existing results in the output directory, so they could not be opened or compared as the procedure requires; no claim is made about them. Unattended run: nothing requiring consent (no forms, uploads, logins) was attempted.