Identity & Access

What camera-based identity and access actually does

A badge reader knows a card was presented. It does not know who presented it. A lost card grants access until someone notices and reports it. A cloned card grants access indefinitely. A badge handed to a colleague for convenience — tailgating at the door, or sharing for a shift — is invisible to a credential-only system.

Camera-based identity closes that gap. The camera sees the person, not just the credential. It knows whether the face presenting the badge matches the enrolled profile. It detects a second person following through a door before it closes. It maintains an audit trail that answers not just 'was card 4471 used at door 3 at 08:14' but 'who was that, what did they look like, and where did they go next.'

That capability — connecting a physical presence to an identity, in real time, on your hardware, with a visual record — is what this system provides.

The three things camera-based access does that credentials cannot

1. Confirms identity, not just possession

A keycard confirms someone has the card. A camera confirms the person holding the card is the person enrolled against that card number. These are different facts. In a controlled environment — a data centre, a clinical ward, a server room — the difference is material.

2. Detects what credentials are blind to

Tailgating — a second person following an authorised entry through a door before it closes — is invisible to a card reader. Camera-based access detects it in real time. The same applies to a person entering a zone they're authorised to access but then moving into an adjacent zone they're not — something a door-mounted reader never sees once the door is open.

3. Produces evidence, not just logs

A card-swipe log tells you a credential was used. A camera-based access system tells you who used it, what they looked like, and where they went. For compliance purposes — a data centre audit, a clinical incident review, a regulatory inspection — that difference is significant.

The identity methods and when each is appropriate

'Camera-based access' is not one thing. Several different underlying approaches produce different capabilities, different accuracy levels, and critically — different legal and compliance profiles. Understanding which method is being used matters for your legal team as much as your security team.

Camera-based identity and access is a hardware-software co-design problem with a compliance layer that is as load-bearing as the ML stack. The choice of identity method, the model architecture, the enrolment database design, the threshold configuration, and the data residency architecture are all decisions that interact — and all of them have legal as well as technical consequences.

This section covers the full stack: hardware topology, the inference pipeline per identity method, the enrolment and gallery architecture, performance targets, deployment topologies, and the hard problems we plan for.

Hardware topology

Deploying a camera-based access system is day one. Maintaining it accurately, keeping the enrolment database current, managing retention obligations, handling subject access requests, and keeping models performing as your population and environment change — that is the operational reality. This section covers how Athena AI-built identity and access systems live in production.

Enrolment management

Initial enrolment

Bulk enrolment tools accept identity records from HR systems (name, role, zone permissions, email) and prompt each individual through a self-service or operator-assisted capture flow. Image quality is enforced at capture — minimum resolution, frontal angle, eye openness, no occlusion. Low-quality submissions are rejected with specific feedback. Consent records are captured and timestamped at enrolment for face recognition deployments.

Ongoing maintenance

Enrolment records expire at a configurable interval (typically 12–24 months for workplace deployments). Expiry triggers a re-enrolment prompt — not automatic access revocation, but an alert to the identity owner and their manager. Repeated false rejects against an enrolled identity trigger a re-enrolment prompt automatically. Leavers are removed via HR system integration or manual revocation — removal is immediate and propagates to edge node caches within a configurable window.

Visitor management

Temporary enrolments with defined access windows and automatic expiry. Visitor enrolment captured at reception — a single frontal capture, associated with the visit record, expires at end-of-day or end-of-visit. No consent gap: visitors are informed and consent obtained as part of reception check-in. Visitor templates are deleted automatically at expiry.

Compliance operations