Todo

• able to create valid state machines
  • first, each transition and state is visible (generated by graph)
  • prune by merging with custom automata learning algo
• post-corpus-parsing, use myers diffing algorithm to remove subtle variants (session cookies, etc, but keep param names, etc)
• generate dataset and classifier for negative samples
• look into hybrid learning (look into a few techniques)
  • [] revised merging algorithm must have confidence score of promoting blue -> red, human verifies this, or sets limits to confidence to promote
• look into developing a burp fuzzer extension

Optional

• look into developing a burp http tagging extension for paths, easier to find paths during parsing - do we actually need this though? the existing path heuristic works ok

Done

• look into human in the loop hybridized learning (https://arxiv.org/pdf/1707.09430)
• check if adding sub-paths improves graph (gives it more accepting states)

Compilation and Run Command

mvn -q exec:java -Dexec.mainClass=webmapper.WebMapper -Dexec.args="corpus/ctfreqs"