Gannet Optimization Algorithm with Attention Enhanced Deep Learning for Intrusions Detecting in IoT

As a new paradigm, the Internet of Things (IoT) incorporates the Internet and physical objects belonging to many different areas including human health, home automation, environmental monitoring, and industrial processes.In our daily activities, it deepens the presence of Internet-connected devices bringing, along with various challenges, and benefits related to security problems.Over the years, Intrusion Detection Systems (IDS) have been instrumental in the security of information systems and networks.However, applying classical IDS approaches to IoT is challenging owing to its specific features such as specific protocol stacks, standards, and constrainedresource devices.Deep learning (DL), specifically recurrent methods, is successfully executed in the IoT forensics analysis but the main problem of recurrent DL algorithms is that they cannot be parallelized and struggle with long traffic sequences.This study concentrates on the development of a gannet optimization algorithm with an Attention Deep Learning based Intrusion Detection (GOAADL-ID) approach.The presented GOAADL-ID technique mainly intends to boost security from the IoT platform via the detection of intrusions.In the GOAADL-ID system, Z-score normalization is primarily employed to scale the input data.Next, the GOAADL-ID approach applies GOA to elect an optimal subset of features.Moreover, the intrusions are identified by the use of attention long short-term memory (ALSTM).Furthermore, the hyper parameters of the ALSTM system were effectually chosen by the design of the Northern Goshawk optimization (NGO) algorithm.The experimental values of the GOAADL-ID methodology are studied on a benchmark IoT database.The obtained outcomes stated that the GOAADL-ID system results in better performance over other compared approaches.