icy17 / mfpair Goto Github PK

PreProcess:
    1.craw api doc, and generate API structs. Refer to the `example_data/libpcap.json` and `example_data/struct_libpcap.json` for the data format
    2.split sentence and preprocess doc by running `Preprocess/split_sentence.py`
        i.You need to change `in_dir`, `out_dir`, `libs` in this code. `in_dir` should contain the crawed api doc of `libs`. `out_dir` contains results of each lib.
        ii.cmd: `python3 ./split_sentence.py`
Identify Malloc/Free API:
    1.Use finetune-QA model by running `QA/run_QA.py`
        i.You need to change `in_path`, `api_path`, `access_token` in this code. 
        `in_path` contains some sentences, which are the results of `PreProcess-2`. 
        `api_path` contains the apis of target lib. Refer to the `QA/API-list` for the data format.
        `access_token` is the user token of huggingface.
        ii.cmd: `python3 ./run_QA.py <model_name> <out_dir>`
        `model_name` is the name of finetune-QA model.
        `out_dir` contains results of `run_QA.py`
    2.parse the result of model by running `QA/MF_identify.py`
        i.You need to change `gt_api_path` and `target_libs`. `gt_api_path` contains the apis of target libs. `target_libs` is a list of libs which will be parsed.
        ii.cmd: `python3 ./MF_identify.py <in_dir>`
        `in_dir` is the `out_dir` of the previous step(result dir of run_QA.py).
        iii. this step will generate results in `in_dir/final_api`
Generate MF Pair:
    1.Run `Pair/match_api_index.py` to identify the API parameters corresponding to the malloc/free object.
        i.You need to change `in_path` and `out_path`.`in_path` is the result of previous step(MF_identify.py). `out_path` contains the result(parameter index).
        ii.cmd: `python3 ./match_api_index.py`
    2.Run `Pair/match_pair.py` to match malloc-free pairs.
        i.You need to change `in_path` and `out_path`.`in_path` is the result of previous step. `out_path` contains the result.
        ii.cmd: `python3 ./match_pair.py`