Energy-efficient deep learning-based intrusion detection system for edge computing: a novel DNN-KDQ model

Abstract

The proliferation of the Internet of Things (IoT) and edge computing technologies has expanded the attack surface for cybercriminals, emphasizing the need for robust and efficient cybersecurity measures. Intrusion detection systems (IDS) are essential defenses; however, deploying traditional IDS solutions on edge devices remains challenging due to limited computational, memory, and energy resources. This research proposes an energy-efficient IDS framework based on a modified Deep Neural Network with Knowledge Distillation and Quantization (DNN-KDQ) to address these challenges. The CICIDS2017 dataset was preprocessed to extract energy-centric features, and adaptive sampling and model compression techniques were applied to optimize performance. The proposed DNN-KDQ model achieves a prediction test accuracy of 99.43%, reduces model size from 196.77 KB to 20.18 KB, and achieves an inference time of 0.07 ms per sample in real-time scenarios. These results demonstrate the feasibility of deploying high-accuracy, low-latency IDS models on resource-constrained edge devices, contributing to more scalable and energy-efficient cybersecurity solutions for modern network infrastructures.