Hybrid Classifier with Improved Normalization Technique for Attack Detection in IoT

Nowadays, individuals are integrating the Internet of Things (IoT) into their everyday lives. IoT applications, especially smart wearables are widely used these days. IoT security undoubtedly draws research attention due to its extensive and significant applications. However, IoT systems are exposed to numerous kinds of attacks. Hence, to improve IoT security, detecting the attacks is necessary, which is accomplished via IDS systems. Thereby, a new approach is developed for attack detection in an IoT environment. Initially, pre-processing is performed by means of an Improved Log Transform (ILT)-based Normalization model. Then, attack detection is performed using a hybrid model, LinkNet and Bidirectional Gated Recurrent Unit (Bi-GRU) models. Finally, the soft voting method is adopted for combining the outcomes from LinkNet and BiGRU to provide the final detected outcomes.