Multi-layer Event Analytic Method of Adaptive Software Orienting at Uncertain Environments

Abstract

Adaptive software can autonomously adjust system parameters, structure, and behavior in response to possible anomalies. Event analysis, which is responsible for data perception and predictive analysis of the entire system, is the foundation of the entire adaptive process. However, as the size of software systems increases dramatically, the systems interact more closely with each other. This makes it difficult to guarantee the system in terms of recognition accuracy. And in a more dynamic and uncertain environment, event analysis has a lag, which is difficult to meet the tasks with high real-time requirements. Most of the existing event analysis methods are based on rule-based reasoning and ontology reasoning methods, which cannot well meet the requirements of event recognition accuracy and timeliness. To solve the above problems, this paper proposes an adaptive software multi-level event analysis method for uncertain environments. First, we use fuzzing to eliminate noise interference and other problems in the data collection process. Second, an event recognition method based on “anomaly detection and hybrid inference” is established to solve the inaccurate event mapping relationship and reduce the system overhead. Finally, an event prediction method based on dynamic multi-fault trees and multi-valued decision diagrams is established to meet the system's real-time requirements. The experimental results show that the proposed method can effectively ensure the effective detection of system anomalies and the accurate identification of abnormal events, and realize the reliable analysis of anomalies under uncertain environments.