← Breach analysis by the Dual Vulnerabilities framework

Norsk Hydro ransomware attack (LockerGoga group, 2019)

2019 · Norway · C - Manufacturing · Ransomware attack after a months-long undetected intrusion

Analysed 2026-11-07 by Hermes

Open this case in the tool →

What happened

Ransomware attack after a months-long undetected intrusion

Dual vulnerability type

Expertise to Execution

Evidence coverage: 4/6 stages, weight 33/60

Decisive stage: detection (clearest evidence) · Alternative: Strategy to Execution

Why this classification

(1) The expectation at Norsk Hydro, a globally operating aluminium and energy producer with production-critical IT and OT, was material and largely in place: antivirus that did detect the implant within days, trusted/cloud backups used for the rebuild, and a crisis setup that decided within hours - and Hydro's own later disclosure acknowledges a detection shortfall, stating it had 'reorganized our security team to better detect and respond to cyber incidents'. (2) In practice a LockerGoga trojan was implanted in December 2018 through an attachment from a spoofed trusted customer; antivirus found it days after implantation but the foothold survived, and the attacker held roughly three months of undetected access, escalating through Mimikatz, renamed PsExec, Metasploit and Cobalt Strike to Windows Active Directory and pushing LockerGoga from Hydro's own domain controllers; the intrusion surfaced only at detonation on 19 March 2019 (Hydro's own timeline: the attackers 'move through and up hierarchy undetected'), forcing a precautionary global shutdown of about 23,000 PCs and 3,000 servers in 40 countries, encryption of HMI/OT systems, destruction of some backups, weeks of manual operations, a total cost Hydro estimated at around NOK 800 million and full recovery over a year. The response and disclosure were prompt and transparent (no ransom, press conference and market notification on 19 March, Kripos report on 20 March), so the response stage shows alignment, not a gap. (3) The classification rests on the detection stage, the clearest-documented one: the remedies and tools existed and even fired, yet the alert and the precursor observables did not lead to eviction for about 90 days - knowledge sat in the technical column and the chain broke in realising it, with no source documenting a doctrine or leadership decision that was deferred or ignored, so this is Expertise to Execution rather than a doctrinal-column gap. The entry vector (spearphishing attachment) is documented but not load-bearing for the type; the containment gaps (AD-wide escalation, one-domain blast radius into OT, online backups) and the recovery gap (destroyed, non-isolated backups) point to the same type. (4) RM34 (Expertise to Execution lens): detection tooling was present and raised a signal, but insufficiently configured, staffed and acted upon to surface a roughly three-month compromise and an undetected climb through the directory, while INL assessed nearly all precursor observables as perceivable (causal role: Obscuring). RM41 (Expertise to Execution lens): an escalation path from standard-user accounts to domain-wide control through credential harvesting and AD weaknesses remained open for about 90 days, showing the specialist privilege-tiering and directory-hardening work was never executed across the estate (causal role: Amplifying). RM30 (Expertise to Execution lens): one compromise reached the whole estate - 23,000 PCs and 3,000 servers across 40 countries, including HMI/OT systems - and forced a precautionary global shutdown, showing containment capability was materially insufficient rather than merely breached (causal role: Amplifying). RM31 (Expertise to Execution lens): the attacker destroyed some backups before recovery, so those copies were connected rather than isolated, and recovery had to be rebuilt from the remaining trusted/cloud backups over more than a year (causal role: Constraining).

Evidence by stage

Risk markers

Sources

Notes

Case file is a cyberdefensemagazine row with blind=Yes; the three leads in source_url were all opened and all concern the Norsk Hydro LockerGoga attack of March 2019, matching the stated victim. Source caveats: the Europol newsroom page for the October 2021 action renders only a script shell, and the politiet.no release did not load, so the attribution is cited via BleepingComputer's reporting of the Europol announcement and Kripos statement; the actor was unattributed at the time of the attack and attribution is not a DV claim. Minor timeline differences between sources: CyOTE places the first ransom note at the Magnor plant at about 22:45 UTC on 18 March, Hydro's own timeline reports detection via unusual server activity just after midnight CET on 19 March, and early reporting says the attack began in a US plant on Monday 18 March; the foothold dates from December 2018. Cost figures are consistent: Hydro's own estimate around NOK 800 million; CyOTE/Microsoft report over USD 71 million. The Wired article (a case-file lead) adds expert context that LockerGoga does not self-propagate and was pushed via Active Directory/Group Policy; it was only partially extractable behind the article preview. No consent-required actions were needed; no leak sites or stolen data were used.