Research And Implementation Of Automatic Alarm Cluster On Defect Testing System

Static defect detection technology is popular due to using a variety of static analysis methods to find potential defect in the program.But the detection results usually have a large number of alarms and high false positive,which requires the developer to invest a lot of time and effort to manually confirm.Reducing the false alarm rate of results and improving the efficiency of manual confirmation are two effective ways to improve the practicality of static defect detection technology.This paper proposes a method of the automatic alarm cluster based on DTS(short for Defect Testing System)based on a large amount of analysis of defect detection results.Based on the characteristics of the alarm and the data sources of the variables,this paper first proposes four relationship among the alarms:equivalence,dominance,similarity and control.Then this paper analyzes the nature and trigger reason of each relationship,and gives the judgment strategy.Finally,based on DTS,the algorithm of automatic cluster of alarms is given by using the feature function of variables.The method of the automatic alarm cluster has been implemented in DTS.Based on the result of alarm cluster,if an alarm is surely confirmed as a false postive,then the corresponding one who has a equivalence relationship with this false postive can also be confirmed as a false positive,the one who has a similar relationship with the false postive can be most likely judged as a false positive,and it is also established on the contrary.The above method can reduce the amount of the alarm confirmation.By analyzing the testing results of five open-source projects,this method can improve the manual confirmation efficiency of about 60%and provide strong support for the detection of large-scale software by the static defect detection tool.