FedGuardSandbox: A Proposed Secured Cloud Sandbox Based Federated Learning Framework

The conventional approach to Machine Learning fails to provide comprehensive data security, leading to the emergence of Federated Learning as a promising alternative to centralization without jeopardizing the privacy of sensitive data. In defiance of introducing a decentralized approach for ensuring sophisticated security of sensitive data, Federated Learning infrastructure is susceptible to vulnerabilities and adversarial attacks even after executing numerous encryption policies and security standards, such as integrating Differential Privacy, Homomorphic Encryption, Secure Multi-Party Computing, and Blockchain. This paper illustrates potential security flaws and state-of-the-art attacks on Federated Learning and explores the inadequacies of current security measures. Then, this paper proposes an advanced security solution of Federated Learning while incorporating a hybrid malware analysis based automated cloud sandbox to safeguard sensitive data. Through extensive analysis, this paper depicts the significance of security in Federated Learning and outlines viable options to influence future research opportunities.