A hybrid AI based framework for enhancing security in satellite based IoT networks using high performance computing architecture

IoT device security has become a major concern as a result of the rapid expansion of the Internet of Things (IoT) and the growing adoption of cloud computing for central monitoring and management. In order to provide centrally managed services each IoT device have to connect to their respective High-Performance Computing (HPC) clouds. The ever increasing deployment of Internet of Things (IoT) devices linked to HPC clouds use various medium such as wired and wireless. The security challenges increases further when these devices communicate over satellite links. This Satellite-Based IoT-HPC Cloud architecture poses new security concerns which exacerbates this problem. An intrusion detection technology integrated in the central cloud is suggested as a potential remedy to monitor and detect aberrant activity within the network in order to allay these worries. However, the enormous amounts of data generated by IoT devices and their constrained computing power dose not allow to implement IDS techniques at source and renders towards typical central Intrusion Detection Systems (IDS) ineffectiveness. Moreover, to protect these systems, powerful intrusion detection techniques are required due to the inherent vulnerabilities of IoT devices and the possible hazards during data transmission.During the course of literature survey it is revealed that the research work has been done to detect few types of attacks by using the old school model of IDS. The computational expensiveness in terms of processing time is also an important parameter to be considered. This work introduces a novel Embedded Hybrid Deep Learning-based intrusion detection technique (EHID) based on embedded hybrid deep learning that is created specifically for IoT devices linked to HPC clouds via satellite connectivity. Two Deep Learning (DL) algorithms are integrated in the proposed method to improve detection abilities with decent accuracy while considering the processing time and number of trainable parameters to detect 14 types of threats. It segregates among the normal and attack traffic. We also modify the conventional IDS approach and propose architectural change to harness the processing power of central server of cloud. This hybrid approach effectively detects threats by harnessing the computing power available at HPC cloud along with leveraging the power of AI. Additionally, the proposed system enables real-time monitoring and detection of intrusions while providing monitoring and management services through HPC using IoT-generated data. Experiments on Edge-IIoTset Cyber Security Dataset of IoT & IIoT indicate improved detection accuracy, reduced false positives, and efficient computational performance.