The purpose of this thesis is to investigate the implementation of digital authorization for computer systems, specifically how to implement an efficient and secure authorization engine that uses policies and attributes to calculate authorization. The architecture for the authorization engine is discussed, the efficiency of the engine is characterized by various tests, and the security model is reviewed against other presently existing models. The resulting efforts showed an increase in efficiency of almost two orders of magnitude, along with a reduction in the amount of processing power required to run the engine. The main focus of the work is how to provide precise, performant authorization using policies and attributes in a way that does not require the authorization engine to break domain boundaries by directly accessing data stores. Specifically, by pushing attributes from source domains into the authorization service, domains do not require the authorization service to have access to the data stores of the domain, nor is the authorization service required to have credentials to access data via APIs. This model also allows for a significant reduction in data motion as attributes need only be sent over the network once (when the attribute changes) as opposed to every time that the engine needs the attribute or every time that an attribute cache needs to be refreshed, resulting in a more secure way to store attributes for authorization purposes.