ID Analyzer
← Voltar ao Blog
On-PremiseID FortSecurityCompliance

Air-Gapped Identity Verification With ID Fort: Deploying KYC On-Premise for Banks, Governments and Regulated Networks

ID Analyzer TeamNov 22, 20266 min de leitura
Air-Gapped Identity Verification With ID Fort: Deploying KYC On-Premise for Banks, Governments and Regulated Networks

Some organizations cannot send identity documents to a third-party cloud. Central banks, defense contractors, government agencies and institutions under strict data-residency rules often run sensitive workloads on networks with no route to the public internet.

Identity verification still has to happen in these environments. Passports get scanned at counters, employees get vetted, and citizens enroll in services. This post covers how to deploy verification inside an isolated network with ID Fort, ID Analyzer's on-premise offering. It focuses on the work that happens after you have decided to go on-premise: architecture, data flow, updates and day-to-day operations.

What "Air-Gapped" Actually Means for KYC

The term gets used loosely, so define it before you plan anything. Most regulated deployments fall into one of three patterns:

  • Fully isolated. No network path to the outside world. Software and data move in only on approved media through a controlled transfer process.
  • Segmented with controlled egress. The verification servers sit in a restricted zone. Outbound traffic is blocked except through tightly governed channels, if any exist.
  • Private cloud or data-center hosted. Everything runs on infrastructure you own or control, and no personal data leaves it. External connectivity for maintenance may be allowed.

Your security team's definition determines how you handle updates, licensing, monitoring and screening data. Write it down and get sign-off before deployment begins.

Reference Architecture

A typical isolated deployment has four layers.

1. Capture Points

These are document scanners, branch kiosks, staff workstations or internal web apps that collect ID images and selfies. They should talk only to your internal verification endpoint and never to an external service.

2. Application Layer

Your onboarding or case-management system receives captures and calls the verification service. Keep this layer responsible for business logic: which checks to run, which thresholds to apply, and when to route a case to manual review.

3. Verification Layer (ID Fort)

This layer runs on your servers and handles the identity work:

  • Document OCR across a library of 3,000+ document formats from 190+ countries
  • MRZ and barcode reading
  • Document authentication and anti-forgery checks
  • Biometric face match between the document portrait and a live capture
  • Liveness detection

4. Storage and Audit

This covers verification results, images you are required to retain, and logs. In an isolated network you own the whole retention lifecycle, including encryption at rest, access control and deletion schedules.

Mapping the Data Flow

Before go-live, draw every hop a document image takes, from the moment it is captured to the moment it is deleted. For each hop, record:

  • Which system holds the data
  • Whether it is encrypted in transit and at rest
  • Who or what can read it
  • How long it persists

Isolated environments sometimes give a false sense of safety. A network with no internet access can still leak through overly broad internal access, unencrypted file shares or debug logs that capture full document images. The data-flow map is the artifact that auditors and internal security reviewers will ask for first.

Heads up

Check your logging configuration early. Verbose application or proxy logs can quietly store extracted names, document numbers and dates of birth in plain text, outside your retention controls.

Handling Updates Without Internet Access

Running offline does not remove the need for change. Countries reissue passports and driver licenses, and fraud techniques keep shifting. An isolated deployment needs a defined process for bringing in new software and document-model updates.

A workable process usually includes these steps:

  1. Receive the update package through your approved channel.
  2. Verify integrity using checksums or signatures before it crosses into the secure zone.
  3. Stage it in a non-production environment that mirrors production.
  4. Regression test against an internal set of sample documents that represent your real traffic.
  5. Promote to production during a change window, with a rollback plan.

Agree on an update cadence with your security and change-management teams. If updates sit in a transfer queue for months, document coverage and fraud detection will fall behind what is in circulation.

AML and PEP Screening in Isolated Networks

Document and biometric checks can run entirely on local infrastructure. AML/PEP screening depends on watchlists that change frequently, so it needs a deliberate design decision in an isolated network. The common options are:

  • Importing list data on a schedule through the same controlled transfer process used for software updates
  • Running screening in a separate, less restricted zone that receives only the minimum data needed, such as name, date of birth and nationality, without document images
  • Using a hybrid design in which identity verification stays fully on-premise and screening is handled under a separate risk assessment

Whichever option you choose, document how fresh the list data is and how quickly updates reach production. Regulators will want to know whether a customer screened today was checked against current lists.

Capacity and Resilience Planning

Cloud services absorb traffic spikes for you. On-premise, that responsibility is yours. Plan for:

  • Peak load, such as month-end, enrollment drives or branch openings, rather than average daily volume
  • Redundancy, with at least two verification nodes behind an internal load balancer for production
  • Failover behavior, so you know what capture points do when the verification service is unavailable (queue, retry or fall back to manual review)
  • Hardware lifecycle, so server refreshes and OS patching are scheduled into the same change process as software updates

Load-test with realistic image sizes and check combinations. A request that includes OCR, authentication, face match and liveness puts a different load on the system than an OCR-only request.

Pre-Go-Live Checklist

  • Network isolation model defined and approved
  • Data-flow map completed and reviewed by security
  • Encryption at rest and in transit confirmed on every hop
  • Logging reviewed for accidental PII capture
  • Update intake, verification and staging process documented
  • AML/PEP screening approach decided and documented
  • Regression sample set assembled from representative documents
  • Load test completed at expected peak volume
  • Failover and manual-review fallback tested
  • Retention and deletion schedules configured

Vendor Assurance Still Matters

Running software on your own servers does not remove the need to trust the vendor that builds it. Ask how the software is developed, tested and released, and how security issues are communicated. ID Analyzer is ISO 27001 certified, which gives your risk team a recognized framework to assess against. Your internal controls still govern everything that happens once the software is inside your perimeter.

Comece a verificar

Pronto para verificar seu primeiro documento de identidade?

Créditos de teste gratuitos no cadastro — sem necessidade de cartão.

  • Sem Cartão de Crédito

  • Créditos de Teste Grátis no Registro