Correlation-Aware Hybrid Feature Selection for Efficient Machine Learning-Based Intrusion Detection in Smart Home IoT Networks

With the emergence of smart homes based on the Internet of Things, network-based attacks have become more prevalent and require effective and lightweight intrusion detection (ID). In this study, a Correlation-Aware Hybrid Feature Selection (CAHFS) framework for smart home IDS is proposed. The proposed approach fuses correlation analysis and Mutual Information (MI) and Random Forest (RF) feature importance to eliminate redundant and less important features and keep only those features that contribute most to the classification. To evaluate the different feature-selection strategies, four machine-learning models were tested: Logistic Regression (LR), Decision Tree (DT), Random Forest (RF) and HistGradientBoosting. Among the classifiers, HistGradientBoosting provides the most stable results out of the assessed models. The proposed CAHFS feature set, using only 16 features, yielded 91% accuracy, 91% precision, 90% recall, 90% F1-score, and a ROC-AUC of 0.96. It was also able to attain an accuracy level of about 95% specificity with an overall error rate of 8.9%, which provides a good balance between detecting the attacks and avoiding false alarms. The comparison with other feature-selection methods demonstrated that CAHFS managed to decrease the number of features from 22 to 16, with the same or similar classification results. Overall, the findings indicate that CAHFS can offer a more compact and informative feature set to be used in machine-learning-based intrusion detection in smart-home networks, which may benefit networks with limited computational resources.

@article{Alqhtani2026,
  title     = {Correlation-Aware Hybrid Feature Selection for Efficient Machine Learning-Based Intrusion Detection in Smart Home IoT Networks},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {17},
  number    = {9},
  year      = {2026},
  publisher = {The Science and Information Organization},
  author    = {Samar M. Alqhtani},
  doi       = {10.14569/IJACSA.2026.0170949},
  url       = {https://doi.org/10.14569/IJACSA.2026.0170949}
}