← Breach analysis by the Dual Vulnerabilities framework

United Airlines Data Theft (suspected APT19/Deep Panda, 2015)

2015 · United States · H - Transport and storage · Data theft via network intrusion (espionage, flight manifests stolen)

Analysed 2026-11-07 by Hermes

Open this case in the tool →

What happened

Data theft via network intrusion (espionage, flight manifests stolen)

Dual vulnerability type

Strategy to Leadership

Evidence coverage: 2/6 stages, weight 14/60

Decisive stage: response (tie-break: post-breach stage) · Alternative: Expertise to Execution

Why this classification

(1) Strategic expectation: United, the world's second-largest airline and one of the largest carriers of US government travellers, publicly asserted a security posture - customers assured that their personal information is secure, top advisors and best practices used, notification requirements honoured if warranted - and in May 2015 even launched an industry-first bug bounty, a strategic artefact signalling security intent. (2) In practice: an intrusion tied to the China-backed group behind the OPM and Anthem thefts was present for months - its C2 domain united-airlines.net was registered in April 2014 - while United detected an incursion only in May or early June 2015, with detection helped by digital signatures the OPM investigators handed to a list of possible private-sector victims; flight manifests with passenger identities and travel data (reportedly also M&A strategy information) were copied out, the airline still could not say exactly which data was removed, and leadership's response to the Bloomberg report was denial - these reports are based on pure speculation - with no self-initiated disclosure and no customer notification. The entry vector is undisclosed and not load-bearing; recovery was not in play. (3) The classification rests on the response/disclosure stage (tie-break: post-breach stage, beating the detection stage): the expectation that was not met is doctrinal - the security posture and customer-data assurance leadership itself communicated - and the chain broke at the decision level, not in operations: leadership knew of an active breach investigation and still told customers their information was secure, kept the incident undisclosed (the media disclosed it), and left affected parties uninformed. That is a gap within the doctrinal column - espoused strategic intent diverging from enacted leadership behaviour - i.e. Strategy to Leadership, not a know-doing failure of a technical remedy. The material-column gap (months-long undetected dwell) points secondarily to Expertise to Execution; the unresolved exfiltration scope is a forensic-readiness deficiency described under RM33, not a scored stage gap. (4) RM39 (Strategy to Leadership lens): leadership's disclosure behaviour diverged from its own assurances and notification commitment - the breach surfaced only via media on 29-07-2015, United called the reports pure speculation and never notified customers while its own investigation was ongoing (causal role: Constraining). RM44 (Strategy to Leadership lens): leadership presented to customers a security posture - personal information secure, best practices in place - that the documented months-long, externally-detected intrusion with unresolved data-loss scope materially contradicted (causal role: Predisposing/Obscuring). RM34 (Strategy to Leadership lens): leadership oversight did not ensure meaningful detection readiness - the intrusion ran since at least April 2014 and was found only in May/June 2015 with indicators supplied by OPM investigators (causal role: Obscuring). RM33 (Strategy to Leadership lens): forensic readiness was not in place as part of the incident and legal strategy - at disclosure the airline still could not determine exactly which data was removed, so the loss could not be scoped (causal role: Obscuring).

Evidence by stage

Risk markers

Sources

Notes

EuRepoC row; the case file carries no related_rows. Disclosure came from media (source_disclosure) and attribution is media-based (FireEye's estimate of 10+ victims reaches us only via Bloomberg quoting the firm); no regulator, court or victim document exists, so all incident-specific facts rest on secondary reporting quoting unnamed officials, plus the victim's own quoted statement. The third source_url in the case file (formiche.net Treverton interview, 2024) was opened and contains no incident-specific facts for this row; the classification does not rest on it. Year kept at 2015 per the case file's start_date (01.05.2015, the detection window); C2-domain evidence shows the intrusion infrastructure dates from April 2014. No public follow-up confirms the breach - United never acknowledged it - so the classification of the concealment rests on its own denial statements as quoted. The contemporaneous bug-bounty episode (researcher-reported API flaw unpatched ~6 months) is a separate incident at the same victim and was deliberately not substituted or used to score this row. Reporting also notes manifests alone would likely not trigger state breach-notification duties, which contextualises but does not neutralise the non-disclosure.