Computing dynamic clusters

Dugerdil, Philippe ; Jossi, Sébastien

In: ISEC’09 : proceedings of the 2009 India Software Engineering Conference, Pune, India, February 23-26, 2009. New York : Association for Computing Machinery, 2009, p. 65-73

When trying to reverse engineer software, execution trace analysis is increasingly used. Though, by using this technique we are quickly faced with an enormous amount of data that we must process. While many solutions have been proposed that consist of summarizing, filtering or compressing the trace, the lossless techniques are seldom able to cope with millions of events. Then, we developed a... Plus

Ajouter à la liste personnelle
    Summary
    When trying to reverse engineer software, execution trace analysis is increasingly used. Though, by using this technique we are quickly faced with an enormous amount of data that we must process. While many solutions have been proposed that consist of summarizing, filtering or compressing the trace, the lossless techniques are seldom able to cope with millions of events. Then, we developed a dynamic clustering technique, based on the segmentation of the execution trace that can losslessly process such a large quantity of data. In order to compute the clusters of classes we use a maximal clique computing algorithm. After having presented our technology we show experimental results highlighting that it is robust with respect to the segmentation parameters. Finally we present the tool we developed to compute dynamic clusters from execution traces.