Optimized Hybrid Intrusion Detection in IoT Through Machine Learning Algorithm Implementation

Résumé fourni par la source

As the sum of IoT strategies increases, accurate intrusion detection by IDS becomes a bigger challenge. In this training, several leading ML (machine learning) procedures are applied to the UNSW-NB15 data which consists of records from$\text{5 4 0, 0 4 4}$computer network activities. Some of the algorithms being compared are PLA (Perceptron Learning Algorithm), LR (Logistic Regression), NN (Neural Network), DT (Decision Tree), Voting, Bagging of PLA, AdaBoost and RF (Random Forest). Model performance was checked by accuracy, precision, recall and$\text{F 1}$-score. Out of all the algorithms, Random Forest demonstrated the best results with$\text{8 1. 3 \%}$accuracy, 72.45% precision, 68.9% recall and a 70.63% F1-score, exceeding the performance of PLA and LR. At a higher accuracy rate, neural networks were able to offer moderate results. Again, these results demonstrate how Random Forest stands out among ensemble methods for better IoT hybrid intrusion detection. It suggests exploring more about DL (deep learning) to make recognition in IoT environments more dependable.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.