notes

Personal notes
git clone git://git.laack.co/notes.git
Log | Files | Refs

ELFuzz.md (997B)


      1 # ELFuzz
      2 
      3 **Definition:** Efficient input generation via LLM driven synthesis over fuzzer space is an approach to fuzzing that uses an LLM 
      4 
      5 ## How it Works
      6 
      7 ### Steps
      8 
      9 1. Setup a harness
     10 2. Start with a Naive seed fuzzer
     11 3. Evaluate the seed fuzzer(s)
     12 4. Mutate the seed fuzzer with the goal of exploring more of the fuzzer space
     13 5. Add the mutated fuzzers to the lattice structure, discarding weaker ones
     14 6. Select the mutants that cover the max unioned cover set
     15 7. These selections should then be the new seed fuzzers for the next iteration
     16 8. Return to step 3 if all iterations haven't been completed
     17 
     18 ### Mutation
     19 
     20 There are three mutation approaches described:
     21 
     22 1. Splicing
     23     - Combines the prefix of one candidate and the suffix of another, combined with code from the LLM to fill in the middle
     24 2. Completion
     25      - Truncate a candidate at a random position and fill in the rest with the LLM
     26 3. Infilling
     27     - Remove random lines and use an LLM to fill in the missing parts of the code