Resuming fuzzing

[marcinguy]

What is the best practice to resume fuzzing?

Thanks,

[ttdennis]

In standalone mode newly discovered corpus files are stored in the input directory. If you want to resume fuzzing from where the fuzzer last stopped, you would start it with the last seed that was used during the previous fuzzing session (using the --seed parameter).

In AFL mode, AFL pretty much handles everything resumption-related.

[marcinguy]

I have noticed it reads the coverage and corpus from output directory automatically upon subsequent start. But this is something different, right?

Is --seed taking a directory parameter? Or just appending --seed would do it

[ttdennis]

The seed parameter is pretty much just the seed value which is fed to the RNG. You would probably want to set the seed value to the last seed you've seen in the fuzzer output.

E.g., when the last log line contains:
... [seed=795] ...

continue with --seed 795