115581748/cpp4app

Contextual Privacy Policies for Mobile Apps

This repository is for the project Contextual Privacy Policies for Mobile Apps.

🚀 To the best of our knowledge, we are the first to propose a novel framework, i.e., SeePrivacy, to automatically generate contextual privacy policies for mobile apps.

🥳 We propose a novel multi-modal framework designed to automatically generate contextual privacy policies for mobile app. We utilize computer vision techniques and NLP techniques to extract corresponding segments from the privacy policy document.

👏 We also build the benchmark dataset, i.e., Cpp4App, for contextual privacy policies on mobile apps, including 1,200+ CPPs with their corresponding privacy policy segments.

We build a demo website for SeePrivacy: https://cpp4app.github.io/SeePrivacy/

Overview

Dependency

python 3.7

opencv-python==4.6.0.66

openai==0.27.6

paddleocr==2.5.0.3

paddlepaddle==2.3.1

beautifulsoup4==4.11.1

stanza==1.5.0

Folder Structure

• code/SEM: Segment Extraction Module of SeePrivacy

• code/CDM: Context Detection Module of SeePrivacy

• dataset/: Cpp4App Dataset

How to Use

• To run segment extraction module, modify the input privacy policy path in code/SEM/P1_PP_processing.pyand run the file. The output will be saved in txt.

• To run context detection module, modify the input image path in code/CDM/run_batch.py and run the file. The output will be saved in result_classification.

  Note: You may replace the OpenAI API Key in code/CDM/detect_classify/classification.py with your own key to enable GPT-3.5.

Dataset

For a detailed description of our benchmark dataset, please refer to dataset/README.md.