MULTIPLE NETWORKS SECURITY STRATEGIES FOR DATA SHARING USING MS-FL

Abstract

More and more consumers are using multi-party data for federated machine learning to get the models they want as a result of advancements in data science, artificial intelligence, and data transactions. In order to solve real-world problems, academics have put forth a variety of federated learning frameworks. Nonetheless, contemporary federated learning systems still have three problems that need to be fixed: 1) safeguarding privacy; 2) preventing poisoning; and 3) safeguarding participants' interests. This study suggests MS-FL, a unique federated learning architecture based on several security techniques, as a solution to these problems. Data suppliers don't have to worry about data privacy leaks thanks to the framework's algorithms. It can also protect against malicious nodes' poisoning attacks. Lastly, a blockchain protocol is used to guarantee that the interests of all participants are safeguarded. The theoretical derivation demonstrates this framework's efficacy. The algorithm developed in this paper performs better than previous algorithms, according to experimental results.