← Breach analysis by the Dual Vulnerabilities framework

Limburg.net data theft and extortion (Medusa, 2023)

2023 · Belgium · O - Public administration and defence · Data theft and ransomware extortion via a legacy internet-exposed server

Analysed 2026-11-06 by Hermes

Open this case in the tool →

What happened

Data theft and ransomware extortion via a legacy internet-exposed server

Dual vulnerability type

Strategy to Execution

Evidence coverage: 4/6 stages, weight 24/60

Decisive stage: entry (clearest evidence) · Alternative: Expertise to Execution

Why this classification

(1) Limburg.net, the inter-municipal waste company for 43 Flemish municipalities and Diest, had a working security doctrine on paper: an MFA standard on its access points, a standing rule that the old data server was no longer allowed to be connected to the internet, and statutory GDPR notification procedures it states it followed. (2) In practice none of that reached the legacy estate: attackers used an employee's password on a second access point where MFA had not been applied, entered on 17/11/2023, moved to the poorly secured old data server that was still online contrary to the rule, and over weeks copied 383,000 files - 2014/2015 national-register lists, debt-mediation files and pre-2005 merger documents - without the organisation's monitoring noticing; the intrusion was found on 13/12/2023 only when the hackers announced themselves with a $100,000 extortion demand, and the full scope surfaced only through the attacker's leak, journalistic research and a final EY analysis in June 2024 (292,734 victims, plus 555 bank account numbers contradicting the initial public statements). The entry vector is documented but is not what is decisive: it is that the existing MFA coverage and exposure rule were never operationalised on the legacy server and access path. (3) That is a doctrinal-origin, execution-break gap: the standard existed, was resourced and demonstrably worked (MFA blocked the first attempt), yet was organisationally unhinged from the actual architecture - formal artefacts aspirational rather than realised - which is Strategy to Execution. Containment and recovery show the opposite pattern (immediate preventive shutdown, restart from verified recent backups), so the misalignment is specifically policy-to-architecture, not across the board. The classification rests on the entry stage, the only gap whose origin and break are documented without inference; the undetected dwell and unmonitored egress point to Expertise to Execution but rest on inference and do not win the stage. (4) RM22 (Strategy to Execution lens): the authentication and account-lifecycle policy was never converted into enforced coverage - MFA was deployed and blocked the first Medusa attempt, yet a second access point remained outside its scope and an employee password opened the door (causal role: Enabling). RM27 (Strategy to Execution lens): the exposure rule barring the old data server from the internet was a baseline that was simply not applied to the deployed system, which stayed poorly secured and reachable until it was hacked (causal role: Enabling). RM34 (Strategy to Execution lens): monitoring and alert-handling requirements were not operated as intended against the legacy server - a 26-day dwell ended only when the attacker self-disclosed (causal role: Obscuring). RM35 (Strategy to Execution lens): no egress control translated into an operational barrier - 383,000 files left the network undetected and the theft had to be learned from the extortion message and the dark-web publication (causal role: Amplifying).

Evidence by stage

Risk markers

Sources

Notes

EuRepoC row; victim, country and year from the case file (receiver_country Belgium, start_date 13.12.2023). Attribution to Medusa is media-based (eurepoc_extra attribution_type); the victim's own publications opened here do not name Medusa. The detail that the old server was 'no longer allowed to be connected to the internet' comes from Belgian press reporting as relayed by PONT Data&Privacy (secondary) and is not confirmed in the victim's own statements; the victim's own timeline does not mention it. The victim's official timeline PDF gives 17/11/2023 as the hackers' date of entry, while the EuRepoC row dates the incident 13.12.2023 (the discovery day); Year 2023 under either reading. The Gegevensbeschermingsautoriteit announced on 19/11/2024 that it is investigating the data breach (multiple complaints received, outcome to be published); no decision was published as of this analysis, so no regulator finding on the victim's controls exists yet - the Limburg prosecutor was also examining citizen complaints alleging serious negligence in data storage. The initial public account (19/01/2024) that no account numbers were taken was corrected by the final analysis of 25/06/2024; the organisation states it always communicated only what was known at the time and notified the supervision authority of each new aspect. The dataset source_url https://www.limburg.net/nieuws/de-impact-van-de-cyberaanval no longer opens; the January 2024 press release PDF and the current cyberaanval page on limburg.net carry the same content and are cited instead. No related_rows in the case file. Link check: lesoir.be blocks automated access (HTTP 403); the page was opened and read during research.