"Architectural Analysis IDs 78 Specific Risks in Machine-Learning Systems"

A formal risk framework aimed at improving the development of secure machine learning (ML) systems has been developed by researchers at the Berryville Institute of Machine Learning (BIML). The BIML researchers conducted an architectural risk analysis of ML systems, concentrating on highlighting the issues that engineers and developers need to consider in the design of ML systems. BIML researchers' architectural analysis delved into the different components of a typical ML system, including raw data, dataset assembly, and learning algorithms. The data security risks associated with each of the components, such as data poisoning, subtle nudges to an online learning system, and more, were identified and ranked. The identification, ranking, and categorization of these risks can help engineers and developers figure out what security controls need to be implemented to mitigate those risks. This article continues to discuss BIML's architectural risk analysis of ML systems and the importance of securing data when using such systems.  

Dark Reading reports "Architectural Analysis IDs 78 Specific Risks in Machine-Learning Systems"