An IoT-Focused Intrusion Detection System Approach Based on Preprocessing Characterization for Cybersecurity Datasets

Security in IoT networks is currently mandatory, due to the high amount of data that has be handled. These systems are vulnerable several cybersecurity attacks, which increasing number and sophistication. Due this reason, new intrusion detection techniques have developed, being as accurate possible for these scenarios. Intrusion based on machine learning algorithms already shown a performance terms accuracy. This research proposes study evaluation preprocessing traffic categorization neural network algorithm. uses its two benchmark datasets, namely UGR16 UNSW-NB15, one most used KDD99. The were evaluated accordance with scalar normalization functions. All models applied through different sets characteristics composed by four groups features: basic connection features, content characteristics, statistical finally, group traffic-based features direction-based characteristics. objective evaluate using various obtain model. Our proposal shows that, applying techniques, accuracy can enhanced up 45%. specific allows greater accuracy, allowing algorithm correctly classify parameters related attacks.