ViewIntent

ViewInent.zip

Folder

• ./backbone
  • The structure of basic module (e.g. the encoder of yolo and the rpn for rcnn) are stored here.
• ./cfg
  • The configuration of yolo and rcnn are stored here.
• ./params
  • The parameters of our models are stored here. We only provide the parameter of ViewIntent now, the parameters of other baselines may be available
• ./utils
  • Some AI-related tools are stored here.

Files

• ./requirements.txt
• It stores the necessary python module for our project.
• ./detectors.py
  • It contains the model structure of our baselines and the old version of ViewIntent.
  • The latest version of ViewIntent which we have proposed in our paper is stored in ./detectors_extra.py
  • The structure of basic models (e.g. yolo and rcnn) are stored in ./yoloModels.py, ./rcnnModels.py, ./pcModels.py
  • The related tools for model construction are stored in ./detectors_utils.py
• ./pretrain_dataloaders.py
  • It provides the dataloader of our pretrain dataset.
• ./pretrain_rcnn.py
  • It pretrains the image encoder (rcnn encoder in our project) on the pretrain datasets.
• ./pretrain_rcnn_eval.py
  • It evaluates the performance of image encoder on our pretrain task (not necessary for ViewIntent).
• ./dataloaders.py
  • It provides the dataloader of our major dataset.
  • The tools we applied to extra the text information from xml files are implemented in xml_model_api.py
• ./finetune.py
  • It trains ViewIntent and other baselines to identify the intention of apps in specfic scenes.
• ./test.py
  • It evaluates the performance of ViewIntent and other baselines on the intention identification task.
• ./analyse_api_res.py
  • It further judge whether the target apps are privacy violated or not based on the prediction of ViewIntent. 
  • The tools we applied to extra the api are implemented in ./PreproText4Bert.py, TextParser_UI.py, UserInputDetector.py
  • The handcrafted rules we applied to judge the violation are stored in ./rule0918.json

EvaluationDataset.zip

Folder

• ./<CategoryA>
  • The intention of samples inside this folder are identified as CategoryA. 
• ./<CategoryA>_<CategoryB>
  • The intention of samples inside this folder are identified as the CatgoryA and CategoryB.