Flag to AHU — Spesimen registry data-quality issue (signature files belong to the wrong official)
Date: 2026-07-14 · Reporter: OCR/apostille team · Severity: High for any automated signature verification; also affects manual verifikator review.
Summary
In the newer spesimen-file/YYYYMM/… format of the Apostille specimen registry (v1_appapos.spesimen / spesimen_position), a registry entry named after one official frequently carries a signature-specimen file that belongs to a different official — specifically the successor who is attesting on behalf of a retired officer. The registry officer_name and the signature inside the attached file do not match. The older pre-migration format (/uploads/spesimen_migrasi/…) is fine — it holds the officer's own genuine specimen.
This means: fetching an officer's registry specimen and comparing it to a document that officer signed can compare against the wrong person's signature, producing meaningless/misleading results (and, for a naive "best of N specimens" match, actively surfacing false matches).
Concrete evidence — officer "Drs. H. Darsani, M.Si" (Disdukcapil Kabupaten Garut)
Four registry entries share this name. We fetched and visually inspected each attached spesimen_file:
spesimen.id |
Registry officer_name |
File format | Signature actually in the file | Genuine Darsani? |
|---|---|---|---|---|
| 2655 | Drs. H. Darsani, M.Si | pre-migration "CONTOH TANDA TANGAN" sheet | Darsani | ✅ yes |
| 290994327 | Drs. Darsani, M.Si | LAMPIRAN FORMULIR 02 "Keterangan Keabsahan" | Sari Nurlistiana, S.Sos, M.Si (successor) | ❌ no |
| 291063455 | Drs.H.DARSANI,M .Si | LAMPIRAN FORMULIR 01+02 | Sari Nurlistiana (specimens on p.1; p.2 attests about Darsani) | ❌ no |
| 291092585 | Drs. Darsani, M.Si | (earlier finding) | a different official (Galih Yudha Praja) | ❌ no |
The FORMULIR 02 files are attestations ("Keterangan Keabsahan Tanda Tangan & Cap") stating that Darsani validly signed a past document (e.g. Kutipan Akta Kelahiran no. 4457/k/2004), and explicitly note "Sudah Pensiun dan telah diganti Kepala Dinas yang baru" — Darsani has retired. The signature block on those forms is the current official's (Sari Nurlistiana), not Darsani's. So the file is genuinely not Darsani's signature specimen; it only references him by name.
3 of the 4 "Darsani" entries carry the wrong person's signature. Only the old-format entry (2655) is usable.
Why it matters
- Automated signature verification (our AI advisory) compares the document signature against the registry specimen. When the registry file is a successor's signature, the comparison is against the wrong person — the score is noise. In the Darsani case the wrong-person file (291092585) even scored the highest of the four, so a "take the best specimen" strategy would pick a false match.
- Manual verifikator review is misled the same way — a verifikator opening the "Darsani" specimen sees someone else's signature and can't verify against it.
- This looks systemic, not a one-off: the FORMULIR 02 pattern (successor attests retired officer, file carries successor's signature, entry named after the retired officer) is how retirements appear to be recorded.
Requested action (AHU side)
- Treat the FORMULIR 02 "Keterangan Keabsahan" files as attestations about an officer, not as that officer's signature specimen. Either (a) store them under the attesting (current) official's registry entry, or (b) mark them as attestation-type so downstream systems don't use them as the named officer's specimen.
- Where a retired officer's genuine own-specimen exists (e.g. the pre-migration sheet), keep it as the canonical specimen for that officer.
- Ideally expose a field distinguishing "own signature specimen" vs "attestation-by-successor" so consumers can filter.
Our side (no change requested from AHU)
We can detect most of these by OCR-reading the "Nama Pejabat" on the specimen form and comparing it to the registry officer_name; a mismatch flags a contaminated file. Not built yet — flagged separately. Until then, our advisory already treats low scores as "AI cannot judge" (never "different"), so a contaminated specimen degrades to a non-verdict rather than a false negative.