2025-09-11
2026-06-18
2025-12-04
Manuscript received August 29, 2026; accepted September 22, 2026; published November 20, 2026
Abstract—Real-time smart-grid intrusion detection requires models that are fast enough to keep up with operational traffic and accurate enough to safeguard vital control functions. This paper analyses aggregate results for nine intrusion-detection models from one smart-grid cyber range. It is a structured secondary analysis, not a new experiment or a statistical meta-analysis. The reported accuracies range from 98.32% to 100%, and the displayed model-inference times range from zero (after rounding) to 0.056879 ms. The study reproduces simple latency ratios and examines possible three-stage routing scenarios. The accuracy-latency trade-offs of these nine machine-learning and deep learning models, namely, Random Forest (RF), K-Nearest Neighbors (KNN), Extreme Gradient Boosting (XGBoost), Gated Recurrent Unit (GRU), Deep Belief Network (DBN), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Deep Neural Network (DNN), and Long Short-Term Memory (LSTM) were evaluated in an emulated multilayer cyber-physical smart grid intrusion detection system. Among positive-latency models, XGBoost and DNN form the principal non-dominated accuracy-latency frontier: XGBoost reports 99.9% accuracy at 0.000671 ms, while DNN reports 99.49% at 0.001156 ms and the best Layer 3 error profile. The paper identifies what can be inferred from the available tables and specifies the data and validation needed before any deployment recommendation is justified. Keywords—cyber-physical systems, cybersecurity, deep learning, machine learning, smart grid, intrusion detection, pareto frontier, real-time systems Cite: J. E. Efiong, A. Akinwale, E. A. Olajubu, G. A. Aderounmu, "Accuracy‐latency Trade‐offs in Machine and Deep Learning Models for Real‐time Smart Grid Intrusion Detection," International Journal of Smart Grid and Clean Energy, Vol. 15, No. 2, pp. 108-122, 2026. doi: 10.12720/sgce.15.2.108-122 Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).