Comparative Study of Lightweight Encryption Algorithms Leveraging Neural Processing Unit for Artificial Internet of Medical Things

Abstract

The Artificial Internet of Medical Things (AIoMT)enables a new generation of medical devices with real-time data analytics, remote patient monitoring, and tailored medicine. This interconnected landscape also facilitates cyberattacks targeting sensitive and critical patient information.

Cryptography is one Method of ensuring secure data transmission. IoT networks have boosted the concept of lightweight cryptography since IoT devices have limited resources, including power, memory, and batteries. These algorithms are designed to protect data efficiently while utilizing minimal resources.

The research presents a comparative study of lightweight encryption algorithms evaluated by the National Institute of Standards and Technology (NIST) for suitability in securing data on AIoMT devices. Here, we analyze the Functional and Non-Functional characteristics of leading contenders. The value proposition of this research is to address the need to secure critical, sensitive patient information on AIoMT devices.

The evaluation is performed using Raspberry Pi AI Kit, integrated with an M.2 HAT+ board and a Hailo-8L accelerator module; the Method adopted is a systematic literature review. Eight Models adopted AES, PRESENT, MSEA, LEA, XTEA, SIMON, PRINCE, and RECTANGLE; ML models adopted and trained and verified against each of the eight NIST lightweight encryption algorithms and every model assessed with key performance indicators such as precision, recall, F1-score, and accuracy.