How Singapore's War on Duplicate Images in Public Records Reached a Tipping Point

Singapore's government technology agencies are accelerating efforts to eliminate duplicate images from public-sector databases, a problem that has quietly accumulated over two decades of digitisation and now touches everything from HDB flat listings on the Housing Development Board portal to National Library Board archives and the SingPass identity verification system.

The issue matters now because Singapore is simultaneously expanding its Smart Nation infrastructure and tightening data governance rules under the Personal Data Protection Commission's revised advisory guidelines, which came into force in January 2026. Duplicate images are not merely a storage nuisance — they inflate database sizes, slow retrieval times, and in identity-linked systems, can create mismatches that delay transactions for citizens and businesses alike.

How the Problem Built Up

The roots stretch back to the early 2000s, when government agencies began scanning paper records independently of one another. The Urban Redevelopment Authority digitised planning maps and site photographs. The National Heritage Board archived festival images and heritage building surveys. HDB uploaded flat photographs to its resale portal. Each agency built its own repository with its own naming conventions, and cross-agency data standards were inconsistently applied.

By 2019, when the Government Technology Agency — better known as GovTech — conducted an internal audit across selected whole-of-government data stores, the scale of duplication had become significant enough to flag as a structural issue. GovTech does not publish the raw findings of internal audits, but the agency's subsequent investments in a common data infrastructure platform, called the Singapore Government Data Architecture or SGDA, were directly informed by those findings.

Orchard Road's OneStop data centre corridor and the Mapletree Business City campus in Pasir Panjang both host government cloud nodes where legacy image files have been migrated in batches since 2022. The migration process itself surfaced duplicates at scale — the same photograph of a Toa Payoh void deck, for instance, appearing in three separate agencies' records under different file names and metadata tags.

The Cost of Doing Nothing

Storage is not cheap at government-grade security classifications. Cloud storage costs for the Singapore public sector are governed by GovTech's bulk procurement agreements, and while contract values are not publicly itemised, the agency's budget allocation for digital infrastructure rose to S$3.8 billion across the five-year Smart Nation 2.0 roadmap announced in 2023. Even marginal inefficiencies across hundreds of millions of stored files add up.

The Personal Data Protection Commission's updated advisory guidelines also introduced a de facto pressure point: organisations — including public agencies — are now expected to maintain accurate, non-redundant records where personal data is involved. A citizen's photograph appearing multiple times in the SingPass face verification pipeline, for example, carries compliance implications that did not exist under earlier rules.

The National Library Board's digital archive team at the Lee Kong Chian Reference Library on Victoria Street has been running a parallel deduplication exercise on its collection of heritage photographs since mid-2025, using perceptual hashing algorithms that compare images by visual similarity rather than file metadata alone. The technique, widely used in commercial platforms, is relatively new to Singapore's public-sector workflow.

For residents and businesses, the practical effects are already visible in small ways. HDB's resale portal refreshed its image database in March 2026, removing visibly repeated flat photographs that had appeared across multiple listings. Agents in Queenstown and Tampines reported faster load times for listing pages in the weeks after the update.

Going forward, GovTech's SGDA is expected to include a centralised image registry that assigns unique identifiers to photographs at the point of upload, preventing duplicates from entering the system rather than hunting for them after the fact. Agencies will be required to adopt the registry for new uploads from the fourth quarter of 2026. Legacy deduplication — the harder, slower work — is projected to continue in phases through 2028, with the SingPass identity systems prioritised first given their direct link to citizen services and data protection obligations.