Border and Customs Security beyond the Checkpoint

A border is not a line, it is an approach, a checkpoint, and everything either side of it. The hard part was never recognition at the booth. It is correlating what happened between the ports of entry, at three in the morning, across cameras nobody was watching.

The Capability

Recognition, Correlation, Collaboration

IREX brings facial and vehicle recognition to border and customs estates, with real-time alerts on flagged individuals and vehicles, and instant search across the whole camera network rather than only at the point of inspection. Plates are checked against stolen, uninsured and unregistered records through integrations agreed at design time.

Because the platform is multi-border and multi-modal on one deployment, a crossing, a seaport and an airport can share intelligence under role-based access, with each agency seeing what its jurisdiction entitles it to. This is the same federated pattern described on the Government as a Service page.

Anti-trafficking is a first-class use case here, not a footnote. IREX works with the National Child Protection Task Force, the Anti-Human Trafficking Intelligence Initiative, Million Kids and Zero Trafficking, and missing-person watchlists can run on border cameras like any other list.

One distinction governs everything on this page. IREX reports events and watchlist matches. It does not determine status. There is no biometric signature for a person's immigration status, and the platform makes no attempt to infer one. What it reports is that someone crossed a fence line at 02:14, that a vehicle on a customs watchlist entered lane four, that a boat closed on a stretch of coast outside any port. What that means, and what happens next, is an officer's decision under their own legal authority. The Case ID and audit trail record who made it.

Coverage

Across the Corridor

Watchlist Recognition

Persons of interest and missing persons matched against enrolled lists in real time, with a recorded lawful basis on every search. Only pre-registered people are recognized.

See Ethical Facial Recognition

Vehicle Intelligence

Plates plus make, model, color and type, checked against external databases, with route reconstruction across the approach.

See License Plate Recognition

Interagency Collaboration

One platform, role-scoped by agency and jurisdiction, with secure messaging and evidence upload between them.

See Sover Communications

Perimeter and Green Border

People or vehicles crossing an approach, a wash or an unofficial crossing point, on cameras placed to see the run of the line rather than one panel of it.

See Perimeter Intrusion Detection

Full Audit Logs

Append-only logging with operator identity and lawful reason, exported daily, readable by oversight bodies.

See Ethical AI & Case ID

Thermal Alongside the Cameras

A thermal imager is registered as a sensor in its own right and its events reach the same screens as the camera events. It is not an input to the analytics: every IREX module works on visible-light video.

See Perimeter Intrusion Detection

PTZ under Operator Control

An officer pans, tilts and zooms a PTZ camera from the video player, and each use is logged as an event. Nothing points itself at an alarm, and the intrusion detectors run on fixed cameras.

See Video Management

Events into Your Own Systems

Any alarm can be pushed with its frames to any system your agency can reach over HTTP, and a REST API covers lists, monitors and users. Sigur is documented for access control and Netping as an external device; an NVR or other ONVIF device registers by IP and hands over its streams.

See Real-Time Alerts & Evidence

Scenarios No Catalog ShipsBeta

A border has conditions specific to its terrain and its traffic. StreamVLM™ turns a sentence into a detector with no dataset and no training cycle: a small boat closing on a beach, a truck being unloaded at the roadside, a vehicle stopped where vehicles do not stop. Shipping in beta on selected instances.

The Missions

Four Border Problems, and What Actually Detects Them

  1. 01

    Unauthorized Crossing

    Between the ports of entry, the question is simply whether anyone crossed. The detector watches a fence line, a wash, or a stretch of open ground for people, vehicles, or both, in color low-light at night rather than infrared, and the alert says a crossing happened, where and when. It says nothing about who the person is or what their status may be: the platform has no way to know either, and does not guess.

    Perimeter Intrusion Detection
  2. 02

    Human Trafficking

    This is watchlist work, and the watchlist is the point. IREX matches only pre-registered people, and trafficking victims are one of the named categories the platform exists to find, alongside missing children and wanted suspects. The same corridor search that reconstructs a vehicle's route reconstructs a person's, which is how a sighting at a crossing becomes a direction of travel. IREX partners with the National Child Protection Task Force to apply exactly this to child-protection casework.

    Missing Persons
  3. 03

    Drug Trafficking

    Narcotics move in vehicles, and vehicles are the platform's strongest signal. Plate plus make, model, color and body type are read at every equipped gantry and checked against the lists that matter, so a vehicle of interest is recognized on arrival rather than found afterwards. From one sighting the route reconstructs in both directions in time, across every camera on the corridor. That is the difference between knowing a car crossed and knowing where it went.

    Multi-Camera Object Tracking
  4. 04

    Smuggling and Contraband

    The customs question is usually about behavior rather than identity: a truck that leaves the through route, a vehicle stopped where nothing stops, a transfer at the roadside at an hour with no reason for one. Those are conditions an officer can describe in a sentence, and the inspection lanes themselves are covered by plate and vehicle recognition with an evidence package attached to every read.

Facial Recognition at a Border

The Constraints That Come with the Capability

  • The platform matches against watchlists of pre-registered people only: wanted suspects, missing children, trafficking victims. It cannot recognize or track a traveler who is not on a list, and that limit is enforced in the architecture rather than by policy alone.
  • The database available for real-time biometric identification is restricted in size, deliberately, and facial biometrics are handled as special-category personal data.
  • Every privacy-sensitive action requires a Case ID naming its lawful grounds before it runs: watchlist enrollment, a person search, even viewing live video. The same gate applies to an action taken through the API as to the same action on screen.
  • The log is append-only and exported daily under signature to an external archive, so an ethics committee, an inspector general or a court-appointed auditor can review it without being given access to the platform.
  • IREX is not an identity-verification system. It does not read travel documents, chips or machine-readable zones, and it performs no one-to-one document matching at a booth. What it does is open-set matching against lists you enrolled.

Facts

No Border Reference
IREX has production deployments in eight countries and none of them is a border or customs program. This page describes documented capability, not a reference. The nearest production analogues are an airport and unnamed critical-infrastructure perimeters.
NIST FRVT, March 2021
In its March 2021 submission to the NIST Face Recognition Vendor Test, the algorithm ranked #1 in accuracy among all US companies and top 10 globally of 268 submitted algorithms, on the Border and Kiosk datasets of the FRVT 1:N Identification track. It is a dated result and does not represent a current or continuing ranking.
Visible Light
Every analytics module is specified against visible-light video. Thermal imagers are supported as a registered sensor whose own events are ingested, not as an input to IREX analytics.

FAQ

Can it work across several crossings at once?

Yes. The architecture is carrier-grade and multi-tenant, so one deployment covers multiple crossings and modes with hierarchical role-based access separating what each agency sees. A single instance is proven at 100,000+ cameras.

Connectivity at remote crossings is poor. Does that matter?

It changes the deployment model rather than the capability. An edge server runs analytics locally and sends only events, keeps recording offline when the link drops, and synchronizes automatically on recovery.

How does this help against trafficking?

Missing-person and victim watchlists run on border cameras exactly like any other list, with alerts the moment a person appears. IREX works with the National Child Protection Task Force, the Anti-Human Trafficking Intelligence Initiative, Million Kids and Zero Trafficking, and holds datasets of thousands of missing children.

Can an alarm point a PTZ camera at the target?

Not automatically, and we will not imply otherwise. What is documented is manual control: a PTZ camera is registered with a port, optional ONVIF credentials and a speed multiplier, an operator drives it from the video player, and each use is logged. Presets, tours, auto-tracking and slew-to-cue are not documented platform features. The intrusion detectors themselves run on fixed cameras.

Do you integrate with fence sensors, buried cable or radar?

Not as named integrations, and the honest answer matters here. What is documented is Sigur for access control, Netping as an external device, ONVIF device and NVR registration, a REST API, and an alarm monitor that can push events and frames to any endpoint reachable over HTTP. If your perimeter intrusion detection system accepts an HTTP call, it can be driven from IREX. A specific sensor integration is a scoping conversation, not something to assume.

Who owns the data?

You do. Customers retain 100% ownership and control of their data, and IREX neither owns nor accesses customer video, events, logs, watchlists, or floor plans. On-premises deployment is the data-residency mechanism.

How do we start?

With a pilot: one site, three to five use cases, six to twelve weeks, and two or three measurable success criteria agreed in writing before it begins. Accuracy is benchmarked on your own cameras, and the measured numbers go into the contract.

Start at One Crossing

One crossing, instrumented across the approach rather than only at the booth, is the honest pilot.