Clustering

Workflow

• Run SOME_OTHER_THING3 to generate a ground truth cluster.
• Run mk_command to make a set of commands for GNU parallel to run.
• Run parallel on the generated commands to generate the feature vectors.
• Run SOME_OTHER_THING to reduce all of the feature vectors.
• Run SOME_OTHER_THING2 to cluster the feature vectors into clusters.
• Run cluster_compare on all of the different clusters and the ground truth cluster and record the results.

Example

There's an experiment script that does all of the above. You invoke it like this:

./experiment.sh /media/disk/fuzzing_data/flasm-groundtruth-final.csv flasm-gcc-noopt1 /media/disk/fuzzing_data/flasm-gcc-noopt1/flasm-gcc-noopt1 '/media/disk/fuzzing_data/flasm-gcc-noopt1/src/flasm -d'