Martin SCHMATZ
Dr. Martin Schmatz is a Principal Researcher at IBM Research Europe in Zurich/Switzerland. He studied electrical engineering and received his doctorate from ETH Zurich on the topic of "Noise Parameter Measurements." After joining IBM in 1999, he led research in the areas of I/O link technology and, from 2012, key components of modern IT server systems. Since 2017, he has focused his work on key management systems (KMS) for Cloud applications and, since 2020, in particular on the migration to quantum-secure communication. He holds an MBA, has over 50 scientific publications, and over 100 patents
Session
With the help of readily available OpenSSL call-back functions, it is possible to accurately measure the CPU-time spent in the various stages of a TLS v1.3 session establishment (ClientHello, ServerHello, CertificateVertify, etc). Accurate timings for the computation of the related crypto algorithms used during those stages can then be extracted. While standalone and/or synthetic performance measurements of crypto algorithms are broadly available, the presented approach thanks to in-situ measurements not only characterizes the actual algorithm implementation as used in OpenSSL, but also takes required collateral performance penalties for the use of the algorithms in the TLS v1.3 protocol into account (e.g., memory allocations, data copying, algorithm loading, merging hybrid secrets, etc.). This presentation provides an overview on the measurement techniques and shows detailed measured results with focus on various combinations of ML-KEM and ML-DSA which were introduced in OpenSSL v3.5.