Network Intrusion Detection System for Smart Environments with High performing Novel Feature Selection Approach

The Internet of Things (IoT) has experienced significant growth in data traffic, resulting in increased complexity and dimensionality. Intrusion Detection Systems (IDS) play a vital role in protecting against various types of cyber threats. IoT IDS systems encounter major challenges because of the varied functionality and physical diversity within IoT environments. The vast and ever-changing nature of IoT traffic makes it challenging and impractical to utilize all features and attributes for effective self-protection in IDS. To tackle these challenges, this study presents AN-SFS: A Dynamic and Adaptive Method for Optimal Feature Selection that adjusts to local data statistics. As opposed to standard approaches, AN-SFS modifies its threshold based on current feature subsets and neighboring feature statistics, enhancing its ability to capture subtle relationships and dependencies for more robust and effective selection. We assessed our model using the CSECICIDS-2018 dataset and found it to outperform traditional machine learning classifiers in detection rate, precision, and recall. It achieved an exceptional accuracy rate of 99.9% and a low false positive rate of 0.022, markedly exceeding the performance of current methods.