Perfume (ASE 2014 supplementary content)
This page contains supplementary content for the Perfume paper,
Behavioral Resource-Aware Model Inference, which was published at the
29th IEEE/ACM International Conference on Automated Software Engineering
(ASE14). Learn more at the Perfume main page.
All models referenced below are event-based FSMs, as Perfume currently outputs models in this form. While models shown in the ASE 2014 paper are state-based instead, these models could be trivially converted to state-based models by pushing each node's event type into its outgoing or incoming edges. The former strategy should be used if events in the log were recorded as they were starting, while the latter should be used if events were logged as they finished.
All models referenced below are event-based FSMs, as Perfume currently outputs models in this form. While models shown in the ASE 2014 paper are state-based instead, these models could be trivially converted to state-based models by pushing each node's event type into its outgoing or incoming edges. The former strategy should be used if events in the log were recorded as they were starting, while the latter should be used if events were logged as they finished.
User study
You can view the full
Perfume user study
, which has been collapsed onto one page for convenience.
Modeling TCP
You can download the log
[ txt ]
of TCP packets used as input and the resulting Perfume model
[ png ], which has been
modified so that server-generated events are shaded.
Modeling a real estate website (comparison to BEAR)
