Lightweight IoT Authentication Protocols: A Comparative Review

Abstract

The Internet of Things (IoT) is expected to interconnect more than 75 billion devices worldwide, yet device authenticity remains one of the most pressing unsolved security challenges in the IoT space. Typical IoT nodes have limited computing power, memory, and battery capacity, making traditional public-key-based authentication difficult to implement without compromising either security or resource conservation. This paper presents a structured narrative review and quantitative comparison of lightweight authentication protocols for IoT environments published between 2024 and 2026, spanning seven families: Elliptic Curve Cryptography (ECC)-based, ECC for Radio Frequency Identification (RFID), hash-based, Physical Unclonable Function (PUF)-based, biometric and behavioural, blockchain-assisted, and machine-learning-augmented protocols. The review adds message-level protocol-flow comparisons for representative ECC- and PUF-based schemes, a benchmarking table of published latency, message-size, and energy indicators, and five sector-specific case studies. Reported findings include dynamic-credential ECC schemes reducing communication and computational overhead by more than 37% over prior ECC schemes; PUF-based techniques using machine learning to improve modelling-attack resistance by more than 35% over earlier techniques; blockchain-assisted authentication for fog-enabled IoT; and multi-sector schemes such as SELAP, reducing computation and communication cost to 422 ms and 960 bits respectively, against 548 ms and 2048 bits for the earlier ELWSCAS protocol. Protocols are also examined against ephemeral information leakage, modelling attacks on PUFs, node cloning, and physical tampering. No protocol category is universally optimal; selection depends on a deployment's constraints, threat model, and sector. Research is converging on hybrid designs combining hardware-rooted trust, efficient public-key primitives, decentralised trust, and intelligent anomaly detection.