Hybrid Deep-Generative Adversarial Network Based Intrusion Detection Model for Internet of Things Using Binary Particle Swarm Optimization

Abstract

The applications of internet of things networks extensively increasing which provide ease of data communication among interconnected smart devices. IoT connected with smart devices diverse in a range of fields associated with smart cities, smart-transportation, smart- industrial, healthcare, hospitality etc. The smart devices lack with computational power, energy and inconsistent topology. Due to these factors these are most vulnerable to security attacks which affect the transmission reliability of data between nodes. An IoT network connects heterogeneous devices together and generates high volume of data. To provide security against intrusion attacks, deep neural network (DNN) techniques are adopted to detect malicious attacks. We have proposed on an anomaly Hybrid based deep learning-based approach which is Generative Adversarial Network in accordance with detecting malicious intruders. We designed a distributed IDS controller validated over dataset of NSL-KDD and proven with higher performance in detecting the DDOS Distributed- Denial- of service- attacks. Thus, Experimental Results are calculated with predefined threshold values to detect DDoS-attacks and the resultant proves that HD-GAN model offers better intrusion detection with respect to higher accuracy, recall, precision, f-measure, and lower FPR (False-Positive-Rate).