Developing and managing access control systems is challenging due to the dynamic nature of users, resources, and environments. Recent advancements in machine learning (ML) offer promising solutions for automating the extraction of access control attributes, policy mining, verification, and decision-making. Despite these advancements, the application of ML in access control remains fragmented, resulting in an incomplete understanding of best practices. This work aims to systematize the use of ML in access control by identifying key components where ML can address various access control challenges. We propose a novel taxonomy of ML applications within this domain, highlighting current limitations such as the scarcity of public real-world datasets, the complexities of administering ML-based systems, and the opacity of ML model decisions. Additionally, we outline potential future research directions to guide both new and experienced researchers in effectively integrating ML into access control practices.

A common trait of current access control approaches is the challenging need to engineer abstract and intuitive access control models. This entails designing access control information in the form of roles (RBAC), attributes (ABAC), or relationships (ReBAC) as the case may be, and subsequently, designing access control rules. This framework has its benefits but has significant limitations in the context of modern systems that are dynamic, complex, and large-scale, due to which it is difficult to maintain an accurate access control state in the system for a human administrator. This paper proposes Deep Learning Based Access Control (DLBAC) by leveraging significant advances in deep learning technology as a potential solution to this problem. We envision that DLBAC could complement and, in the long-term, has the potential to even replace, classical access control models with a neural network that reduces the burden of access control model engineering and updates. Without loss of generality, we conduct a thorough investigation of a candidate DLBAC model, called DLBAC_alpha, using both real-world and synthetic datasets. We demonstrate the feasibility of the proposed approach by addressing issues related to accuracy, generalization, and explainability. We also discuss challenges and future research directions.