C2 (3 points)
https://knobit.org/id/task/dso-evaluation-environments/C2
Also as Turtle · JSON-LD · RDF/XML
| dcterms:license | https://creativecommons.org/licenses/by/4.0/ |
|---|---|
| owl:sameAs | https://bitbucket.org/reffa/ontology/id/task/dso-evaluation-environments/C2 |
| rdfs:label | C2 (3 points) en |
| schema:hasPart | dso-evaluation-environments C2: pseudonymised training data labelled anonymised |
| schema:hasPart | dso-evaluation-environments C2: minimisation and storage limitation of training data |
| schema:hasPart | dso-evaluation-environments C2: production card data copied to vendors for an incompatible purpose |
| schema:identifier | C2 |
| schema:position | 2 |
| schema:text | Production data in non-production environments. (a) 1 point: EV-22 is called anonymised but is pseudonymised - the HMAC key sits in the environment's own vault, tokens are consistent, and date of birth, postcode and dispute notes remain, so whoever has the environment can re-link records to people: it is still personal data (GDPR recital 26). The fix keeps the key outside the environment (or uses synthetic data), removes or generalises the quasi-identifiers, and has each refresh approved by the card data owner instead of 'not required'. (b) 0.5 each: data minimisation - EV-22 keeps only the 12 columns the models use (plus keys), not all 46; storage limitation - the indefinite monthly snapshots get a defined retention tied to the purpose, then deletion. (c) 1 point: EV-23 put three months of unmasked production card data into the vendors' tenants to test technology - a purpose not compatible with the one the data was collected for (GDPR Art. 5(1)(b), Art. 6(4)), approved by the sponsor rather than the data owner: the answer replaces it with synthetic or properly de-identified data and has the copies deleted now that the proof of concept ended on 2026-09-12, not kept until the contract decision. en |