Lightweight AI-Driven Intrusion Detection and Energy-Efficient Cryptographic Framework for Sustainable and Resilient IoT Network Security

Abstract

With the number of devices that support the Internet of Things (IoT) exploding, there are new and unprecedented security issues to deal with, while classic intrusion detection and cryptographic techniques are too expensive for computational and energy-constrained environments. This paper introduces a single lightweight framework combining an efficient Intrusion Detection System (IDS) supported by an Artificial Intelligence (AI) algorithm with an energy efficient cryptographic engine to guarantee the sustainability and resilience of the IoT network security. The proposed IDS is also designed with a simple neural structure that utilizes streamlined depthwise separable convolution and attention modules, and has less than 0.5 million model parameters with more than 98% detection accuracy. To ensure data protection, a hybrid encryption approach dynamically switches between lightweight ciphers (SPECK, PRESENT) and optimized elliptic curve cryptography depending on current threat level and energy level of the device. The integrated framework, as shown by experiments, achieved a total energy savings of about 30% over the energy consumption of the conventional framework based on AES, and inference times less than 50ms on edge devices. The system is scalable, energy-efficient and can be real-time deployed in heterogenous IoT environments.