Multi-dimensional framework for context-aware IDS selection in edge-IIoT

This is the authors' abstract. We don't add key points for this paper.

The rapid expansion of Industrial Internet of Things networks has increased the demand for intrusion detection systems that balance accuracy, efficiency, and cost. Most studies have focused on accuracy alone, ignoring practical deployment trade-offs. We propose a multi-dimensional evaluation framework that integrates Pareto Frontier Optimization, Attack Surface Coverage scoring, and Total Cost of Ownership modeling. We evaluate three Intrusion Detection System models on the Edge-IIoT 2022 dataset: Convolutional Neural Network-Gated Recurrent Unit-Long Short-Term Memory with Federated Learning, Decision Tree with Autoencoder, and Long Short-Term Memory-Convolutional Neural Network-Attention. Each model is assessed using normalized metrics combined via a weighted synthesis engine.The results show that the Decision Tree with Autoencoder achieved the highest composite score of 78.85%. LSTM-CNN-Attention delivered a maximum security coverage of 98.16%. Convolutional Neural Network-Gated Recurrent Unit Long Short-Term Memory achieved best accuracy with the lowest five-year cost of $1.385 million. The framework enables context-aware intrusion detection system selection for edge-based IIoT environments.