Edge AI has emerged as a viable alternative to cloud-based video surveillance, offering reduced latency, improved privacy, and enhanced data sovereignty. Nevertheless, many existing solutions remain tightly coupled to proprietary cloud services, limiting transparency, scalability, and forensic reliability. This paper presents a fully open-source, cloud-aware yet cloudindependent architecture for AI-based video surveillance deployed entirely at the edge. The system leverages well-established protocols such as RTSP, MQTT, and HTTP, combined with containerized services and structured JSON event processing. A dual operational model is proposed: a real-time notification pipeline and a forensic-grade evidence preservation pipeline. Experimental deployment on ARM edge hardware demonstrates that forensic guarantees can be achieved with limited overhead, while maintaining scalability and operational simplicity.
Scalable and Secure Edge Video Surveillance without Cloud Dependency: An Open-Source, Forensic-Capable Architecture Based on Standard Protocols
Gentile A. F.
;Greco E.;Lamonaca F.;
2026-01-01
Abstract
Edge AI has emerged as a viable alternative to cloud-based video surveillance, offering reduced latency, improved privacy, and enhanced data sovereignty. Nevertheless, many existing solutions remain tightly coupled to proprietary cloud services, limiting transparency, scalability, and forensic reliability. This paper presents a fully open-source, cloud-aware yet cloudindependent architecture for AI-based video surveillance deployed entirely at the edge. The system leverages well-established protocols such as RTSP, MQTT, and HTTP, combined with containerized services and structured JSON event processing. A dual operational model is proposed: a real-time notification pipeline and a forensic-grade evidence preservation pipeline. Experimental deployment on ARM edge hardware demonstrates that forensic guarantees can be achieved with limited overhead, while maintaining scalability and operational simplicity.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


