Defense in Depth: The Foundation of Traditional Security

Traditional physical security has long relied on the proven strategy of defense in depth — a layered approach where multiple independent security measures work together to deter, detect, delay, and respond to threats. These layers are “independent” in that each operates on its own principles and mechanisms, not relying solely on the success of others. Yet they are also “integrated” into a cohesive system, where each layer complements and reinforces the others to form a stronger, unified defense.

For example, perimeter fences deter and delay, cameras detect, access control systems deny, and security personnel respond — each contributing to overall protection. This approach has served as a cornerstone of security design, ensuring that no single failure compromises the entire system. However, even with these layers in place, cumbersome physical security system technologies often leave personnel playing catch-up. When an incident occurs, security teams must manually analyze alarms, review video footage, and piece together context — an effort that takes time and allows incidents to progress and escalate while responders are still assembling the facts.

Response in depth: The modern evolution

Response in depth represents the next evolution in security operations — a dynamic, real-time enhancement of traditional layered security defenses. While defense in depth focused on preventing or slowing threats through multiple passive layers, response in depth adds active capabilities to not only block threat actions but to demotivate and blunt the drive of hostile individuals — disrupting, halting or even reversing attack progress altogether.

Today’s artificial intelligence (AI)-enabled systems allow information from streaming video content analysis, access control events, intercoms, audio and other environmental sensors, and even human-reported observations to be fused into a shared situational picture — continuously updated and immediately actionable. Instead of individual static defense mechanisms, some waiting to be triggered, the system provides an array of intelligent technological and human response capabilities that operate in parallel, continuously adjusting to changing conditions and delivering real-time interventions tailored to the specific stage and nature of each evolving threat.

In this model, incident detection, classification, and response action planning happen concurrently. A camera identifies unauthorized movement in a restricted zone, and within seconds, the system has verified the activity, alerted human operators, adjusted nearby PTZ camera views, notified on-site personnel, and queued intercom announcements — all in parallel. Human operators remain in control, but with AI speeding up context gathering, analysis, and recommendations, their decisions are better informed and delivered faster.

Response in depth reduces the lag time between awareness and action. It introduces new operational layers that confront and deter bad actors early in the timeline — through tailored interventional actions and real-time system responses — before threats escalate. Most importantly, it shifts security postures from passive monitoring to active, adaptive engagement that changes the trajectory of security incidents while they are still unfolding.

Security industry impacts

A new class of security devices and remote services is emerging because AI can now take advantage of the multitude of technological advancements that in recent years have reached high levels of capability, reliability, and maturity. These advances have all combined to create a watershed moment for physical security, one that is changing the entire landscape of safeguarding. Secure and reliable on-site device wireless communications, high-frame-rate high-resolution cameras, devices that self-enroll in their systems, and smartphone mobile apps have all become baseline capabilities.

Also contributing to this shift are:

• Cloud-native infrastructure, enabling massive scalability, high availability, and real-time data processing across distributed environments
• Edge computing, allowing video and sensor data to be processed locally for instant AI inference and reduced bandwidth demand
• Open APIs and industry standards, making it easier to integrate best-in-class technologies and unify previously siloed systems
• AI-ready GPUs and low-power chipsets, now embedded in cameras and access devices, enabling onboard analytics and localized decision-making
• AI-enabled multi-sensor fusion, combining video, audio, radar, and other sensor inputs for richer, more actionable situational context

Additionally, the proliferation of intelligent consumer devices has fueled a global manufacturing ecosystem now capable of producing affordable, purpose-built, multi-technology products that would have been economically or technically unfeasible just a few years ago. This shift doesn’t just lower costs — it accelerates innovation, allowing new security solutions to be developed and deployed faster than ever before.

Remote safeguarding capabilities 

The rapidly advancing AI-enabled technology landscape is driving the emergence of radically more effective remote safeguarding capabilities — solutions that were previously out of reach even for the world’s largest corporations and highest net-worth individuals. Today, these capabilities are becoming accessible to medium and small businesses and even private residences, allowing property owners to better protect their people and property — at a cost that no longer strains their bottom lines.

Design for response in depth

We have never had security technology capabilities like those available now, and so our design thinking must change to take full advantage of them. For example, two-way audio will play a bigger role in incident response — which we ought to start calling situation response, because we now have greater situation awareness and the ability to respond to threat arrival in earlier stages of action. We need to expand the breadth and depth of situation response planning to enable prevention, early intervention, and even the reversal of attack activity. This is where scenario-based response planning for multiple situations comes into play.

Business property fence climbing scenarios

Scripted AI responses to such situations include decision trees, incorporating both automated technical actions and human-in-the-loop involvement. The first step is to determine the intentions of the fence climber. How to respond will vary depending on the climber’s motives and personal situation. In two of the three scenarios below, the climber has a valid reason to enter the property — and in one, it is a matter of life or death.

• Situation 1: A person is fleeing pursuit from an armed robber, climbs the perimeter fence to escape and/or obtain help, and has no hostile intentions toward the business or its people.
• Situation 2: A thief is climbing the business fence in hopes of finding something to steal on the property or inside its buildings.
• Situation 3: A person climbs the business property fence to retrieve a phone that a disgruntled acquaintance has thrown over the fence.

AI video content analysis detects the climbing attempt once the climber’s feet have left the ground. Talk-down via two-way audio begins, with these responses executed simultaneously:

• Automated AI announcement: “You in the green jacket and hat, this fence protects private business property. Do not climb over it or a guard will call the police, and send them a video of your actions.”
• Automated AI spotlight activation: A pan-tilt-zoom controllable spotlight is focused on the climber.
• Automated talk-down readiness: Video of the scene is displayed for the guard’s situational awareness, a Push-to-Talk button is displayed to enable the guard to talk to the climber via the two-way speaker co-located with the camera, and a Call Police button is displayed in case it is needed.
• Automated police notification readiness: A scripted email containing video clips—including the best view of the climber and the announcement with the climber’s reaction — is readied for sending to the police.

The “private property” statement reinforces the legal boundary, clearly defining the climber’s entry as trespassing — justifying police involvement if the person continues. The combination of the specific announcement, visual spotlight, and video recording should serve as a sufficient deterrent in Situation 2. If the climber lands inside the fence, the response continues:

• Automated AI announcement: “If you need help, wave your hands in the air. If you want me to call the police for you, shake your head up and down several times. If you want to talk to me, just walk toward the spotlight and start talking loudly.”

This next step is designed to distinguish Situation 1 from Situation 3. The on-duty guard —  whether on-site or remote — can initiate a conversation, call the police, or do both. In Situation 1, if the robber begins to climb the fence, the system’s next actions depend on whether the initial climber has reported the pursuit. Should the guard instruct the individual to seek shelter on-site? Could the person be allowed into a building for safety?

Many such questions must be considered in the process of designing intelligent responses to unfolding situations. Situation 1 includes life-safety considerations, while Situation 3 raises public relations concerns. For example, helping someone retrieve their phone using the spotlight could generate positive word of mouth about the business—not only in terms of its security, but also the intelligent and helpful way it responded instead of simply calling the police.

For private residences, scenarios may include flower or statue theft, attempted break-ins, or a child’s ball that ends up over the fence. Intelligent, differentiated responses matter. The yard gate may need to be remotely opened — either by the homeowner via their mobile phone or by a remote monitoring center. Neighbors will appreciate a system that enables something as simple as ball retrieval without triggering alarms or involving law enforcement unnecessarily.

This article doesn’t allow space for further elaboration, but it should be clear that human and AI-guided responses — driven by scenario-based decision trees — can be executed far more quickly and effectively than those supported by traditional security systems alone.

Building on traditional security

Most of our traditional security thinking still applies — with some adjustment, and a lot of expansion — when it comes to scenario-based response design considering emerging AI capabilities. By eliminating the delays of manual analysis, Response in Depth empowers security personnel to act decisively, turning reactive catch-up into proactive control. This is the future of security thinking—where technology not only protects but anticipates and responds with unmatched precision.