How Singapore's Push to Clean Up Duplicate Images in Public Records Went From Backroom IT Problem to Policy Priority

Singapore's public sector is sitting on a sprawling archive of duplicated digital images — passport photographs, property listing photos, identity verification portraits and scanned documents — spread across dozens of agency databases, and the effort to systematically remove and replace them has quietly become one of the more consequential data-hygiene exercises in the city-state's recent administrative history.

The problem did not emerge overnight. For most of the 2010s, government agencies operated largely in silos. The Immigration and Checkpoints Authority processed biometric photographs independently of the Housing and Development Board, which in turn maintained its own visual records for flat applications and resale transactions. The Land Transport Authority kept separate image repositories for driving licences and COE-related documentation. Each system was built to serve its own mandate, and none were designed with cross-agency deduplication in mind.

How the Backlog Built Up

The turning point came with the rollout of Singpass Face Verification, which began expanding across public-facing digital services around 2021. For the first time, a single facial-recognition layer was being applied against records held by multiple ministries simultaneously. What engineers discovered was uncomfortable: the same individual's photograph could appear in three, four or even five separate databases, often in different resolutions, different formats and — critically — at different stages of accuracy. Some images had been scanned from physical documents decades old. Others had been digitised through different vendor pipelines, producing files with inconsistent metadata.

The Government Technology Agency of Singapore, known as GovTech, flagged the issue formally as part of its ongoing work on the Singapore Government Tech Stack. The problem was not merely one of storage inefficiency, though that was a factor — cloud storage costs for the public sector are real and audited annually. The deeper concern was data integrity. When a duplicate image carries a different file hash or a conflicting timestamp from its counterpart in another system, automated verification pipelines can flag false mismatches, slowing down transactions that citizens and businesses rely on.

At the HDB Hub on Toa Payoh Lorong 6, flat buyers have occasionally experienced delays in identity verification steps tied to resale flat applications. Officers at the CPF Board's Tampines Regional Centre have similarly encountered cases where document photo records pulled from legacy archives did not reconcile cleanly with Singpass-linked images. These are not widespread failures, but they are symptoms of the same underlying condition.

The Push Toward Systematic Replacement

GovTech published guidance under its Digital Government Blueprint framework calling for agencies to adopt a canonical image model — meaning one authoritative, deduplicated image per individual per record type, stored centrally and referenced by downstream systems rather than copied into them. The blueprint, which has been updated periodically since its initial 2018 release, set a target of rationalising legacy image records across core identity systems by 2025. That deadline has since been extended into the current financial year, which runs to March 2027.

The practical challenge is significant. Migrating and replacing images is not simply a matter of deleting files. Each image is often embedded in audit trails, legal records and archived transaction logs that cannot be altered retroactively without triggering separate compliance reviews under the Personal Data Protection Act. The PDPA, administered by the Personal Data Protection Commission, requires that any changes to stored personal data be documented and justified — a requirement that adds process weight to what might otherwise be a straightforward database operation.

For residents interacting with government services at counters in Jurong East or through the LifeSG app, the day-to-day impact has been limited. Most duplicate-image issues are resolved before they surface at the user level. But the longer-term stakes are higher: Singapore is positioning itself as a regional hub for AI development and digital identity infrastructure, and the credibility of that pitch depends partly on demonstrating that the city's own data foundations are clean, consistent and auditable.

Agencies have been advised to complete image replacement workflows for high-priority record categories — those touching identity, immigration status and property ownership — before the end of the third quarter of 2026. For citizens, the practical advice is straightforward: keep Singpass profile photographs updated and respond promptly to any agency requests to re-submit identity images, since those requests are often part of the broader deduplication exercise rather than a cause for concern.