On the Latency of Secure MQTT over QUIC vs. TCP: An experimental evaluation

Smart-city services increasingly depend on IoT communication infrastructures that must provide secure connectivity without compromising latency-sensitive operation. This creates an engineering trade-off: security mechanisms protect urban sensing and service platforms, but may also increase communication latency, especially when devices connect under load. QUIC addresses part of this trade-off by integrating secure connection establishment into the transport layer, making it a relevant candidate for MQTT-based smart-city deployments. This study experimentally evaluates MQTT over QUIC and TCP under different security configurations, endpoint platforms, and traffic patterns. We separate three latency-relevant operating conditions: connection establishment, concurrent connection scaling, and sustained streaming in a multi-sensor, edge-to-cloud scenario motivated by smart-city applications. The results indicate that, in the evaluated configurations, the main latency cost of security appears during connection establishment, whereas steady-state encrypted data transmission adds limited overhead across the tested traffic patterns. For single connections, QUIC reduces per-connection latency by up to around 3.2 times compared to TCP with TLS. Under concurrent connection bursts, QUIC shows linear tail-latency growth in the measurements, and under high streaming load, it degrades more gradually than the TCP-based alternatives. The findings identify practical trade-offs among transport protocol choice, cryptographic algorithm, key type, and size, and endpoint computing capability. These trade-offs inform the design of event-driven smart-city services in which devices must establish or maintain secure communication under constrained latency and hardware conditions.