Layers, and cameras on every layer.
Data center physical security is designed in rings. The property line and the anti-climb fence. The vehicle gate and guard booth. The building entrance, the lobby and the mantrap. The loading dock. Then the data hall, the cage and the rack. Each layer has access control, and each layer has cameras. The site security operations center (SOC) or the regional GSOC watches all of it on a video wall, and the guard at the front console is expected to notice anything that does not fit.
Unlike most sites, a data center is quiet in the right way. Few people move around the exterior, and the ones who do are expected: shift technicians, escorted vendors, scheduled deliveries. That makes an unexpected person meaningful — and it makes the audit record just as important as the alert.
What the SOC watches.
- Perimeter fence line. Long runs shot by fixed cameras, often with thermal or low-light imaging at night. The event is a person approaching or following the fence, or a vehicle stopping on the outside road.
- Vehicle gate and guard booth. Every vehicle in, driver interaction with the booth officer, and anything that follows the approved vehicle in.
- Loading dock. Deliveries, equipment arriving for installs and hardware leaving for disposal. The event is dock activity outside a scheduled window, a door left open, or a person on the dock with no truck present.
- Mantraps and building entrances. One badge, one person. The event is two people in a single-person portal, or someone waiting at the exterior door for the next badge holder.
- Generator yards and fuel areas. Remote corners of the property with expensive equipment and few visitors.
The nuisance activity is familiar to any SOC operator: security vehicle patrols passing the same cameras on schedule, rain and fog on exterior domes, insects on IR illuminators at night, and badge-holding staff walking the same routes every shift.
Why the cameras don’t make it easy.
Perimeter fence cameras watch long, narrow strips where a person is small and far away, and IR at night flattens contrast. Exterior lighting is deliberately even, but spider webs and bugs near the illuminator trigger motion constantly. Inside, mantrap cameras are close-range and brief — the event lasts seconds. And a large campus runs so many views that the video wall relies on tours and multiple monitors, so the one camera that matters is often not the one on screen.
How Semaphor supports the SOC.
Semaphor watches the same rendered monitoring screens the SOC uses. It works with existing camera software: no camera credentials, no VMS plugin, no new integration touching the security network. For operators who have spent years hardening that network, that alone matters.
On each view, local person and vehicle detection watches for the things a quiet perimeter should not have. A preliminary alert reaches the operator right away through alert-first review, and an AI scene report follows on the same card: a person walking along the inside of the east fence, no vehicle present, guard review suggested. Camera views that cycle off the wall are tracked as unobserved rather than assumed quiet, so coverage gaps are explicit instead of hidden.
The audit trail is the product.
Data center customers and auditors ask for proof. Semaphor keeps the original image for every processed frame and crop, with an exact SHA-256 hash and perceptual hashes (pHash and dHash), and exports JSON manifests that tie each AI assessment to the frames it was based on. That is the evidence and audit trail a SOC can hand to an investigator without anyone wondering whether the image was altered. Retention is automatic and bounded in guard mode, so evidence is there without growing forever.
For a deeper look at the control that matters most indoors, see our guide to tailgating at access doors. For how Semaphor fits a multi-site operations center, read Semaphor for GSOCs.
What your guards watch
- Perimeter fence runs and the outside road
- Vehicle gate and guard booth
- Loading dock doors and staging area
- Mantraps, lobby and exterior entrances
- Generator yards and fuel storage
What Semaphor surfaces
- A person on or near the fence line
- Vehicles stopping outside the fence or following through the gate
- Dock activity with no delivery present
- More than one person at a single-person portal
- Hashed, exportable evidence for every event
Semaphor is an assistant for the operator, not an access control system. The SOC operator reviews every alert and decides whether to escalate.
See it on your wall.
Start with what Semaphor is. Then book a live demo on your own camera screens, or start a pilot on one SOC console and measure missed events, nuisance alerts and evidence review time against a normal shift.
Industrial campuses share many of the same layers. See monitoring manufacturing plants.
Data center monitoring FAQ
Does Semaphor need access to our camera network?
No. It reads the monitoring screen the SOC already displays. No camera credentials or VMS integration are required.
Can Semaphor evidence be used in an investigation?
Every frame keeps its original image and SHA-256 hash, and manifests export as JSON, so investigators can verify exactly what the AI saw and when.
Can AI inference run on site?
Yes. Detection runs locally, and the vision-language model can run locally or on hosted inference over verified HTTPS with authentication and optional mutual TLS. Contact us to discuss your requirements.