A Descriptive Pilot Study of Performance and Resource Utilization of Lightweight Cryptography Implementations

The rapid proliferation of the Internet of Things (IoT), accompanied by the widespread deployment of embedded systems, has significantly increased the demand for cryptographic mechanisms specifically tailored to resource-constrained environments. In response to these challenges, lightweight cryptographic algorithms have been developed with the aim of minimizing memory consumption, code size, and the computational overhead associated with cryptographic operations. The objective of this study was to experimentally evaluate the performance and resource utilization of selected reference implementations of lightweight authenticated encryption schemes under a controlled, platform-specific execution environment. Three authenticated encryption with associated data (AEAD) implementation artifacts were examined: Ascon-128 v1.2, Elephant (Delirium), and GIFT-COFB, all originating from the NIST Lightweight Cryptography standardization process. The Ascon implementation evaluated in this study corresponds specifically to Ascon-128 v1.2 and should not be confused with the subsequently standardized Ascon-AEAD128 specified in NIST SP 800-232. The measurements were conducted on a general-purpose x86_64 host running Linux under WSL2 and Docker; consequently, the reported values characterize the evaluated implementation artifacts on this platform rather than performance on embedded or IoT hardware. The experimental framework included functional correctness testing, throughput and latency measurements, and assessments of code size, memory utilization, stack usage, and the presence of dynamic memory allocation. The measurements revealed substantial differences among the evaluated implementations in computational performance and resource-related metrics on the tested x86_64 platform. The study is descriptive and of a pilot nature: the primary results come from a single measurement session with a small number of aggregated samples and no replicates in the primary session. In these measurements, the evaluated Ascon-128 v1.2 implementation showed the highest throughput and lowest short-call latency, whereas the evaluated Elephant (Delirium) and GIFT-COFB implementations exhibited smaller code footprints under selected compilation configurations. These observations should be interpreted as characteristics of the tested implementations and execution environment rather than as a general ranking of the underlying constructions for embedded platforms. The results therefore provide a platform-specific reference for analyzing implementation trade-offs and motivate validation on representative microcontroller architectures before deployment-oriented conclusions are drawn.