Badge Access Secures Doors.
AI Secures Everything Else.
A Tier III colocation data center in the Southeast addressed the physical access vectors that a professional penetration test had identified — rooftop intrusion and perimeter boundary breaches that bypassed the badge system entirely — using 34 existing exterior cameras and zero new hardware.
The Penetration Test Found What Badge Access Couldn't See.
Data centers invest heavily in logical access security — multi-factor authentication, biometric verification, mantrap vestibules. These controls govern every door. They govern nothing above the roofline. A professional physical penetration test commissioned by the facility's security team identified three access vectors that bypassed the badge system entirely: a rooftop access ladder reachable from an adjacent HVAC service platform, a perimeter fence section with an unmonitored climb approach, and a maintenance hatch accessible with basic tools that was technically locked but physically vulnerable.
The concern was specific and documented: copper theft, cooling system interference, and in a colocation environment, the potential for physical media extraction from customer equipment bays. A rooftop intrusion that went undetected for 20 minutes in a Tier III facility could represent both physical infrastructure damage and a customer data exposure event with contractual and regulatory implications that extended far beyond the cost of the intrusion itself.
Reactive Surveillance vs. Proactive Safety
Physical Envelope Coverage on Existing Exterior Infrastructure
IntelliSee was connected to all 34 exterior cameras — covering every rooftop access ladder, HVAC equipment platform, perimeter fence approach, and vehicle access checkpoint — without new hardware or network modification. Rooftop intrusion detection was configured as the primary alert class on all cameras with rooftop view; perimeter control on fence-line cameras; vehicle detection on after-hours loading dock approaches.
Alert routing integrated directly with the facility's existing SOC monitoring terminal and the on-duty physical security manager's device. Alert classification included camera location, zone designation, and detection image — giving SOC operators the context to initiate the appropriate response protocol without leaving the monitoring station for a visual confirmation.
North Rooftop Access Ladder. 1:22 AM. Saturday.
At 1:22 AM on a Saturday, Camera 31 detected a person ascending the north rooftop access ladder — one of the three vectors identified in the penetration test. The detection model classified the rooftop zone event and routed an alert to the SOC terminal in five seconds. The on-duty SOC manager confirmed the detection and contacted the physical security response team immediately.
The security team reached the rooftop access point at 1:38 AM — 16 minutes after the detection alert. The individual was a competitive intelligence operative hired to demonstrate rooftop access feasibility. They had reached the parapet but had not yet accessed the HVAC service platform that leads to the equipment bay roof. No infrastructure was accessed. No equipment was damaged. The subsequent security report cited the detection as validating both the pen test finding and the effectiveness of the AI monitoring deployment.
The Physical Gaps Badge Access Left Behind — Closed.
The facility had never had automated rooftop monitoring capability. IntelliSee provided the first real-time alert system covering all identified rooftop access vectors, using cameras already installed for other purposes.
Every exterior camera was transformed from passive recording infrastructure into active detection. No new hardware. No installation window. No tenant notification required.
All three access vectors identified in the physical penetration test were brought under continuous automated detection coverage in a single deployment phase.
Rooftop Intrusion Detection: How the Model Knows the Difference Between a Person and HVAC Movement
Rooftop environments are visually complex: HVAC units vibrate, fans rotate, birds land, and lighting conditions change dramatically across a 24-hour cycle. Standard motion detection in these environments produces a false positive rate that makes it operationally useless. IntelliSee's rooftop intrusion model does not detect motion — it detects the object class 'person' within defined spatial zones that correspond to rooftop surfaces, equipment platforms, and access ladder approaches.
This distinction is the operational difference between a system that generates 40 alerts per night and one that generates zero alerts on quiet nights and one alert when a person actually appears. The model is trained to distinguish a person from the visual background of a rooftop environment specifically — not generic outdoor motion detection repurposed for a roof application. For data center physical security, that precision is the threshold requirement for SOC deployment.
We had cameras on the roof for years. We just had no idea what they were seeing. Now we know within five seconds.
Your Cameras Are Watching
Every Surface of Your Facility.
Are You Listening?
We'll audit your existing exterior camera positions against your physical access vectors and show you exactly where detection coverage closes your remaining blind spots.