A gas storage facility, a substation or a port: large areas that are difficult to secure and are increasingly the focus of potential attacks. Traditional video surveillance quickly reaches its limits in such situations. There are too many images, too little context and too much manual analysis. At the same time, however, demands are rising. This is precisely where SecuSite AI, Materna’s AI-based security platform, comes in.
A white van is parked for hours by the fence of a gas storage facility. Two men in anoraks bearing the words ‘Lower Water Authority’ appear to be photographing technical equipment. Local residents become suspicious, and the police launch an investigation – to no avail. This is what happened in 2025 in Kraak, Mecklenburg. Incidents such as this demonstrate just how real the threat has now become. Espionage attempts, suspected sabotage and hybrid threats are increasingly targeting energy facilities, pipelines, military sites and other sensitive infrastructure.
Gas storage facilities, substations, chemical plants, ports, wind and solar farms, and military sites all have one thing in common: they are large areas containing sensitive infrastructure. They are often enclosed by fences, and sometimes by natural barriers such as coastlines, rivers or topographically inaccessible peripheral areas. Yet even where fencing is in place, it is virtually impossible for staff to monitor large areas without leaving any gaps. This is precisely why security systems are needed that effectively complement human vigilance.
A prerequisite for using SecuSite AI is a digital camera infrastructure with IP-based video data and RTSP streams (the Real-Time Streaming Protocol enables access to IP cameras via the web). The AI-based security platform utilises this existing digital surveillance technology as the foundation for an intelligent, scalable security architecture.
Materna integrates SecuSite AI deeply into existing IT and security environments. A detected incident therefore does not end with a simple warning message on a monitor or a shrill alarm bell. Rather, the alert is processed within the correct operational context – from automated incident detection, through handover to existing control centre or ticketing systems, right through to notification of a security service and audit-proof documentation. The deep system integration required for this is a unique selling point, as it ties in precisely with Materna’s core expertise: connecting complex IT systems in such a way that individual data points are transformed into useful processes.
Powerful NVIDIA technology enables comprehensive operational monitoring that goes far beyond traditional perimeter security. For example, the AI can also identify anomalies or, by analysing movement patterns, detect health issues that may indicate leaks in chemical plants.
For operators, this means: real-time analysis of large volumes of video data, high scalability for complex sites, short response times and industry-proven GPU performance. This combination is particularly crucial in the KRITIS environment. Security systems must operate reliably, with low latency and high availability at all times.
The definition of which behaviour or incidents are considered security-relevant is carried out using natural language. SecuSite AI processes not only rigid rules but also context-sensitive descriptions such as: “Alert me to vehicles that repeatedly move slowly along the perimeter fence”, ‘Notify me if a person is present in Zone C outside regular shift hours’ or ‘Detect if someone appears multiple times at different access points’. The platform then analyses the video footage and describes any anomalies it detects within precisely this defined context. This makes video surveillance significantly more precise and, at the same time, more flexible.
Conventional video surveillance often operates according to fixed rules and motion detection. This leads to high false alarm rates, caused, for example, by animals, shadows or weather conditions. Materna’s platform, on the other hand, analyses scenes contextually. The AI recognises what is considered normal in the respective environment and what is not. The result: greater security whilst reducing the workload on control centres.
This applies not only to people and vehicles on the ground. With appropriate AI training and the necessary hardware, the platform can also detect drones. What’s more: provided it has been trained appropriately, it can distinguish patterns in flight movements and, for example, assess whether a flight is a random overflight or a regular, targeted approach to sensitive areas. This is particularly relevant for operators of critical infrastructure. After all, visual surveillance of the ground alone is no longer sufficient by any means when reconnaissance and disruption can also occur from the air.
Data protection is not optional, particularly in the context of critical infrastructure. For this reason, video data can be processed either on-edge, on-premises or in the cloud. This ensures compliance with regulatory requirements, minimises the attack surface, guarantees maximum performance without cloud latencies and provides full control over data processing. As a result, SecuSite AI can be tailored precisely to security requirements, compliance standards and existing IT strategies.
