← Breach analysis by the Dual Vulnerabilities framework

SNUH patient-data theft (suspected Kimsuky, 2021)

2021 · South Korea · Q - Human health and social work activities · State-sponsored data theft via web shell and stolen credentials

Analysed 2026-11-06 by Hermes

Open this case in the tool →

What happened

State-sponsored data theft via web shell and stolen credentials

Dual vulnerability type

Strategy to Execution

Evidence coverage: 4/6 stages, weight 23/60

Decisive stage: egress (clearest evidence) · Alternative: Expertise to Leadership

Why this classification

(1) Strategic expectation at the victim: as one of Korea's four largest national university hospitals processing resident registration numbers and medical records, SNUH stood under a standing Personal Information Protection Act duty to apply security measures (access control, protection/encryption of sensitive data), and its own records show a database-encryption programme reported complete in February 2017. (2) In practice: a North Korean group entered through a weakness of an internet-facing web server in June 2021 (web shell, Remote Desktop into an idle server, internal shared folders), took personal data of 22,020 people including 10,089 resident registration numbers, and about a year later reused stolen credentials on the pathology-data server and the same path into the employee e-newsletter database, taking treatment records of about 810,000 patients and data of 1,953 staff - roughly 831,000 people in total. The hospital noticed the first intrusion only on July 6, 2021, some 26 days later; after the Ministry of Education - NIS joint investigation demanded fixes in July 2021, its own audit records that weaknesses stayed unremediated for over three months because the hospital kept delegating all security management to its vendor and verified neither the vendor's reports nor the contracts' security terms; the second joint investigation found the data still unencrypted despite the 2017 encryption report; and the public account stayed wrong - the July 12, 2022 notice stated resident registration numbers had not been leaked, which the Personal Information Protection Commission's May 2023 decision contradicts. (3) The entry vector is disclosed but not load-bearing. The classification rests on the egress/data-protection stage, where evidence is clearest and primary: a documented legal and organisational requirement existed before the incident and never materialised in the deployed architecture, which is the diagonal Strategy-to-Execution gap (doctrinal origin, execution break) rather than merely weak security; the regulator's first-ever surcharge on a public institution for safety-measure violations confirms the misalignment, and no source shows leadership knew the data was unencrypted beforehand (only the vendor's attestation that it was encrypted). Secondary patterns sit in detection (capability never resourced in-house before January 2022 - Expertise to Leadership reading) and in response (demanded remediation deferred - Strategy to Leadership reading). (4) RM26 (Strategy to Execution lens): the hospital's data-protection requirement - encryption of resident numbers and personal medical data - was never implemented in the systems where that data sat, which is why plaintext crown-jewel data left the network (causal role: Amplifying). RM21 (Strategy to Execution lens): lessons from the first intrusion and the July 2021 joint investigation - remove the weakness, harden the exposure - did not become durable changes; the audit records them unremediated for months and the same path and credentials were reused for two further thefts (causal role: Perpetuating). RM38 (Strategy to Execution lens): after detection the eradication was incomplete - only the web shell was addressed, the exploited weakness stayed open and the acquired account credentials were not effectively removed, so access returned (causal role: Amplifying). RM13 (Strategy to Execution lens): supplier-security requirements did not translate into controlled third-party operation - security management was entrusted wholesale to one vendor, maintenance contracts omitted security clauses, and the vendor's encryption and remediation reports were never verified (causal role: Enabling).

Evidence by stage

Risk markers

Sources

Notes

Case-file data quality: the name/description figures 831,00 and 6,969 - the figure is approximately 831,000 individuals (about 810,000 patients incl. deceased, 17,000 current and former staff; PIPC: 10,089 resident registration numbers, 22,020 personal-data records, 813,038 treatment records, 1,953 employee records), and 6,969 files is the hospital's own July 2021 estimate, which the later police investigation superseded. The KNPA's May 10, 2023 release itself sits in a Korean government viewer document that could not be read directly, so it is cited here via the Korean security-media reproduction of that release; the KNPA report attributes the intrusion to a North Korean group without naming one - the Kimsuky name comes from the July 2021 parliamentarian briefing and expert analysis, hence 'suspected' in the Actor. The dark-web account figures in the inews24 piece are reported only as reported; no leak site or dumped data was accessed, per the research policy. The PIPC plenary decision of 2023-05-10 is cited from established Korean security-media coverage of the decision (boannews, dailysecu, doctorsnews); the PIPC's own release page was not opened and is therefore not cited. No related rows: a search over all case files in cases-run500 found the only 'Seoul National University Hospital' file is this one. Cited pages all opened during this analysis (2026-11-06); the case file's UPI link could not be opened and is not cited.