TLS Protocol Analysis Using IoTST—An IoT Benchmark Based on Scheduler Traces

Abstract

The Internet of Things (IoT) envisions billions of everyday objects sharing information. As new devices, applications and communication protocols are proposed for the IoT context, their evaluation, comparison, tuning and optimization become crucial and raise the need for a proper benchmark. While edge computing aims to provide network efficiency by distributed computing, this article moves towards sensor nodes in order to explore efficiency in the local processing performed by IoT devices. We present IoTST, a benchmark based on per-processor synchronized stack traces with the isolation and precise determination of the introduced overhead. It produces comparable detailed results and assists in determining the configuration that has the best processing operating point so that energy efficiency can also be considered. On benchmarking applications which involve network communication, the results can be influenced by the constant changes that occur in the state of the network. In order to circumvent such problems, different considerations or assumptions were used in the generalization experiments and the comparison to similar studies. To present IoTST usage on a real problem, we implemented it on a commercial off the-shelf (COTS) device and benchmarked a communication protocol, producing comparable results that are unaffected by the current network state. We evaluated different Transport-Layer Security (TLS) 1.3 handshake cipher suites at different frequencies and with various numbers of cores. Among other results, we could determine that the selection of a specific suite (Curve25519 and RSA) can improve the computation latency by up to four times over the worst suite candidate (P-256 and ECDSA), while both providing the same security level (128 bits).