Contents I. Introduction
By moving feature extractors from fast to slow gadgets, model migration may minimize training costs and speed up model convergence throughout federated learning on IoT devices. This allows for more sustainable computing. Nevertheless, federated learning systems aim for efficiency and dependability, yet dishonest or unmotivated devices could move the false models or refuse to share models for their own gain [1]. As more and more businesses go to the cloud to house the massive amounts of data generated by their Internet of Things (IoT) strategies, data auditing takes on greater significance. The reason for this is because customers are able to verify the veracity of the data they are outsourced. However, most existing data auditing methods fall short when it comes to guaranteeing data integrity and meeting the security needs of practical multimedia facilities. In addition, customers do not get their money back quickly when their cloud service provider (CSP) messes with their outsourced data since there is no fair arbitration process in place [2]. There are benefits and drawbacks to the extensive adoption of IoT technology. So that devices can communicate with one another securely and hackers can't have access to sensitive information, the Internet of Things needs a robust and trustworthy security framework. With the help of a detection system, several additional measures may be taken to ensure the security of the Internet of Things. Many effective intrusion detection systems (ID have been developed alongside the advancements in machine learning and deep learning [3]. A lack of sufficient Identity Management has been revealed by the fast growth of linked entities induced by the popularity of the Internet of Things (IoT) as well as Industrial Internet of Things (IIoT). Any disruption to the traditional central authority-dependent IdM system might have far-reaching consequences [4]. A deluge of dispersed data has been produced by the fast expansion of the IoT. The quality of services provided and the rate of innovation may both be improved by sharing this data. Traditional approaches to data exchange, on the other hand, often make use of intermediaries, which increases the likelihood of privacy breaches and single-point failures. It is also difficult to ensure fair payment among data users and suppliers using these conventional sharing techniques since they do not include a secure transaction mechanism to cover the costs of data sharing [5]. Recent years have seen widespread adoption of the internet of things (IoT), which is finding use in many sectors, including healthcare. Some issues, however, do arise during the creation and implementation of IoT data analysis methodologies. These include insufficient training data, limited resources, a lack of a centralized framework, security concerns, and privacy concerns. The proliferation of blockchain technology, on the other hand, offers a decentralized foundation [6]. Supply Chain Management (SCM) is the practice of regulating the movement of goods and services from an organization by attending to each step in the production process, from sourcing raw materials and components to shipping the completed goods to the customer. The proliferation of IoT-enabled smart city applications—from smart grids and residences to smart supply chains and healthcare—has recently captured the public's interest. In order to treat patients quickly, experts are now thinking about the smart healthcare system's function as a PES [7]. With the increasing need for ocean sensing and exploitation, it is essential to establish a Marine Internet of Things (IoT) system capable of collecting and analyzing vast amounts of data about the marine environment. Building a Marine IoT system isn't without its challenges. Some of them include dealing with enormous amounts of heterogeneous data, limited transmission bandwidth underwater, significant heterogeneity, and imbalanced computation.-data-sensing devices' resource burden [8].
