Distributed Software Interactive Behavior Analysis Based on Knowledge Fusion

A novel online analysis method for software Interactive Behavior (IBs) is presented in the paper. Scenario-sensitive method is adopted to model complicated IBs among third party software entities. By fusing real-time self-experience and pervious experience based on knowledge, the creditability of interactive entities is computed automatically. Multi-Entity Bayesian Network (MEBN) tool is adopted to construct reusable domain “knowledge fragments”. If current scenario is similar to pervious one, then pervious one is reused; if there is no similar scenario, evidences gained from monitoring and pervious experience are fused to construct behavior model for this scenario. The combination of large and small granularity knowledge reuse improves analysis efficiency of IBs.