Abstract :
The information processors in a decentralized computing system must trust each other enough to be mutually supportive, yet they must also protect themselves to maintain autonomy. In a decentralized system, data security is especially important because the effects of compromise or sabotage can be so wide-ranging. The trusted function is an ad hoc solution to a problem with present data security models. This "consistency" problem, never previously addressed in a formal manner, is aggravated in the decentralized processing setting. The paper examines the consistency problem and proposes an addition to existing security models to address the problem. Using the model, the impact of the trusted function on secure, decentralized-processing system design is assessed.
