2016 · South Korea · C - Manufacturing · Data theft via network intrusion (state-linked espionage)
Analysed 2026-11-06 by Hermes
Open this case in the tool →Data theft via network intrusion (state-linked espionage)
Strategy to Execution
Evidence coverage: 2/6 stages, weight 11/60
Decisive stage: entry (clearest evidence) · Alternative: Expertise to Execution
(1) Strategic expectation: DSME is the shipbuilder of South Korea's Aegis destroyers and submarines, so a structural expectation sits at the strategic level that the business intranet carrying that classified design work is insulated from the public internet - an architecture standard for a defence prime, not an optional hardening step. (2) In practice: in April 2016 intruders reached DSME's database and took about 40,000 documents, around 60 of them classified military documents (blueprints, construction technology, weapons-system data); defense sources said the company was found vulnerable because it had failed to properly insulate its business intranet from the public internet; the intrusion was found by a Ministry of Defence cybercrime division, not by DSME itself, and the company publicly knew nothing of its own breach until a lawmaker disclosed it on 31 October 2017. The technical entry vector was never disclosed and is not load-bearing for this classification. (3) The gap is doctrinal-to-execution: the expectation (network insulation for classified work) originated at the strategic/architectural level, and the chain broke in realising it - the deployed network simply did not implement the separation, leaving crown-jewel design data on an internet-reachable business network. No source documents a specific policy artefact being drafted, a pentest warning or a deferred budget, so it is not Expertise to Leadership or Strategy to Leadership; nor is it a known patch or remedy specialists sat on (Expertise to Execution) - the failure is an unimplemented architecture standard, matching the method-spec's 'entry can be Strategy to Execution' counter-example. The detection stage shows a secondary Expertise to Execution pattern (material monitoring work that did not identify the intrusion), but it rests only on inference from who discovered the breach and therefore cannot be the decisive stage; the entry stage is decisive on clearest evidence, the only stage whose origin and break are source-documented. (4) RM27 (Strategy to Execution lens): the exposure baseline for a classified-work network - insulation from the public internet - was not applied to DSME's deployed business network, leaving the blueprint database reachable from outside (causal role: Enabling). RM34 (Strategy to Execution lens): monitoring and detection requirements were not operated as intended - the intrusion sat undetected by the victim until a Ministry of Defence investigation division found it, and the company was unaware of its own breach a year and a half later (causal role: Obscuring).
RM27 — Configuration hardening / external exposureRM34 — Monitoring / detection / alert-handling gapAttribution: 'almost 100 percent certain' per lawmaker Kyung Dae-soo, based on the investigative team's method-similarity assessment; no formal public government attribution statement - dataset attribution_type is a direct statement in a media report; North Korea denies involvement. Scale figures (40,000 documents, ~60 classified) trace to Dong-A Ilbo reporting as relayed by Reuters/The Japan Times; investigators did not disclose the classification levels. Dataset caveat: eurepoc source_disclosure says 'Incident disclosed by victim', but Reuters quotes a company spokeswoman who was unaware of the issue until the lawmaker's disclosure - the disclosure ran through the parliamentary briefing, not the company; recorded here as a data-quality remark. The insulation finding rests on a single Dow Jones article citing anonymous defense sources. Recurrence (context, not evidence for this row): the Korea JoongAng Daily reported in 2023 that a different North Korean organisation hacked DSME again (2022-2023), so this is a separate later incident, not an M14/RM21 predecessor of the 2016 breach. Recovery and egress stages not in play / not documented as gaps as recorded in the stage evidence. Unattended run: no source needed consent; no leak-site or attacker-site material consulted. Link check: reuters.com blocks automated access (HTTP 401); the page was opened and read during research. Link check: japantimes.co.jp blocks automated access (HTTP 403); the page was opened and read during research.