Avaliação do impacto do TLS na performance da comunicação MQTT usando mosquitto

Abstract

The MQTT protocol is widely used in Internet of Things (IoT) applications due to its lightweight design and efficiency, features that are crucial for devices with limited com putational resources. However, MQTT does not provide robust native security mecha nisms, making the useofTransport Layer Security (TLS) a commonapproachtoensure confidentiality, authenticity and message integrity. Although TLS enhances communi cation security, it may introduce computational overhead, particularly during the hand shake and continuous data encryption processes. This work investigated the impact of TLS usage on the performance of the Mosquitto broker by comparing equivalent scenarios with and without encryption, under different message loads and numbers of clients, measuring latency, throughput, CPU usage and memory consumption. Results showed that the TLS impact on throughput is negligible under moderate loads, remain ing below 1% up to 75 simultaneous clients, but becomes significant under high loads, reaching an overhead of 9.9% at 100 clients and 12.8% at 110 clients, with instability signs starting at approximately 85 clients. Regarding latency, TLS introduced an over head of 6 to 7% for smaller message volumes, converging with the non-TLS scenario at larger volumes. CPU and memory usage showed low overhead in absolute terms, with nearly equivalent memory consumption between both scenarios (159MB without TLS and 162MB with TLS). These results provide quantitative evidence to support the understanding of the trade-off between security and performance in MQTT-based sys tems.