Genie: An Inference Engine with Applications to Vulnerability Analysis

摘要

An expert system is a program that mimics the performance of a human expert in some decision making process. We are building expert systems to deal with vulnerability analysis and have developed a general-purpose mechanism (called an 'inference engine') with which to emulate the behavior of human experts. The project is called 'Genie', and this report offers a presentation of its design, its implementation, and its usage. Genie is primarily a goal-driven (backward chaining) engine, but whenever it makes a deduction it is used wherever possible in data-driven mode (forward chaining). Production rules are written in simple everyday English; a preprocessor in Genie converts these rules into a frame-based knowledge representation scheme for efficient computation. We treat several facets of the interface between Genie and its human users, including Genie's facilities for explaining and justifying its behavior. Finally, we at the first application of Genie by the Army in a Turbine Engine Expert System and present some of the additions and improvements we plan to implement. An inference engine shell will aid in the development of future expert systems, while the expert systems themselves will make the vulnerability analysts' jobs much easier.