Distributed computing in multi-agent systems: a survey of decentralized machine learning approaches

At present, there is a pressing need for data scientists and academic researchers to devise advanced machine learning and artificial intelligence-driven systems that can effectively tackle forthcoming security challenges. This is due to the rapid advancement in processing capabilities of distributed client devices, as well as mounting apprehensions about the exposure of sensitive user data. This particular methodology renders the constituents of a decentralized machine learning paradigm private and allocates them to a diverse array of remote client apparatuses. Following this, a primary controller clusters the extracted machine learning data. This process effectively transforms a centrally controlled machine learning procedure into a distributed one, leading to significant potential benefits. Regrettably, the implementation of a progressive strategy for networked machine learning poses additional obstacles to the preservation of user data confidentiality and cyber-security apprehensions. The present endeavor aims to scrutinize the plausible hazards that decentralized machine learning may engender for the security and confidentiality of user data, as well as their well-being, through the lens of data transmission thresholds. These risks are contingent upon the salient stages of a machine learning model. The aforementioned procedures include: (i) establishing thresholds for raw data prior to processing; (ii) establishing thresholds for the learning framework; (iii) establishing thresholds for the obtained information; and (iv) establishing thresholds for the provisional outcome. We conducted a comprehensive analysis of state-of-the-art hacking methodologies, evaluating their potential vulnerabilities at every stage of the transaction, and subsequently deliberated on practical solutions to address these issues. Ultimately, the survey culminates in an exposition of the obstacles and prospects that confront researchers in this intricate realm.