![[Pasted image 20230715152910.png]] > According to the IEEE standard [19], the lifecycle（[[IEEE Standard Classification for Software Anomalies]]） of change-induced incidents can be summarized as shown in Fig. 2. A change can introduce a defect to systems or activate a dormant defect existing in systems. A defect is activated to a fault when encountered during system execution. A failure occurs when the fault is propagated to the service interface and causes the service delivered by the system to deviate from the correct service [20]. A failure further cause an incident when it brings down services and has adverse effects on customers. When a fault occurs, the system behaviors usually manifest anomalies. Anomaly detection algorithms can be used to detect anomalies in the system as early as possible after an anomaly appears to avoid incidents. Compared to the lifecycle of non-change-induced incidents, the difference is that there is a change that can introduce defects or activate faults in the lifecycle of change-induced incidents. （出典：[[2023__ICSE__An Empirical Study on Change-induced Incidents of Online Service Systems|Wu+, ICSE-SEIP2023]]） - [20]\: [[2004__TDSC__Basic Concepts and Taxonomy of Dependable and Secure Computing]]