Real-Time IoT Network Intrusion Identification-A Hyperparameter Tuned Machine Learning Approach

Ensuring network security is crucial for protecting confidential data and avoiding unapproved entry into computer networks. We present an efficient approach for identifying network intrusions in this research using a combination of feature engineering, various machine learning classifiers and hyperparameter optimization. The proposed methodology involves preprocessing the network traffic data using principal component analysis PCA to reduce dimensionality, SMOTE to address class imbalance. Subsequently, multiple classifiers are trained on the preprocessed data to detect intrusions in real-time network traffic. Through extensive experimentation and evaluation, it has been observed that machine learning algorithms, including XGBoost, Random Forest (RF) and K- Nearest Neighbours (KNN) outperform the other classification algorithm for accurately identifying network intrusions. By leveraging the capabilities of hyperparameter optimization, aims to further improve the efficiency and effectiveness of network intrusion detection systems, thereby enhancing overall cybersecurity posture. The accuracy of the hyper-optimized RF is 99.6 % and AUC-ROC of 99.99 %, the highest accuracy achieved in comparison to machine learning algorithms. These results provide valuable insights into developing robust intrusion detection systems capable of mitigating evolving cyber threats.