Debloater output sample set

[ameily]

We need a decent sized corpus of debloated binaries to test against that fully exercise DIFFER. In general, our sample set should include:

• A sample that reads from stdin, similar to head
• A sample that reads an input file, similar to cat

We should have output from several delobaters and be able to trigger failures for missing features.

[ameily]

During today's call, it was also mentioned that we could perform analysis on different release versions of a binary. For example, if a feature was added to a command line application within a release, evaluate the original and the updated binary.

[ameily]

We can potentially use output from the "Debloating Tradeoffs" paper

• Artifacts: https://github.com/qixin5/debloating_study
• Data with debloated samples: https://gitlab.com/anony22/debaug_release

[ameily]

Reviewing the debloated samples above, it may be difficult to use them because we don't know the inputs used during delobating or what was actually debloated. The output is the debloated code that we need to compile. So, these may not be ideal since it would require a good amount of work to get working.

[ameily]

I believe this is done. We've added the benchmark programs from the spreadsheet and I believe GT is on the hook for eventually providing debloated versions of these.