API-Risk-Assessment-Framework

This project quantified the risk of the endpoint of each API based on security and data sovereignty markers. In this repository, we have included a well-annotated python script for (1) the data preprocessing and feature engineering, along with the (2) machine learning pipeline. These can be found in the src folder.

This project is part of UBC MDS' capstone project 2022 where the contributors collaborated with TeejLab.

Proposal

Our proposal can be found here.

Final Report

Our final report can be found here.

Technical Report

For a high level summary of our project and to understand our decision making choices, please refer to the technical report. We have included links to the relevant scripts and notebooks for easy navigation within the repository.

Top-level Directory Layout

.
├── data                    # Data files
│   ├── preprocessed           # Preprocessed Dataset
│   └── raw                    # Raw Dataset
├── docs                    # Final Proposal and Report
│   ├── proposal_book          # Final Proposal
│   └── report_book            # Final Report
├── model                   # Model file
├── notebooks               # Jupyter Notebook files 
│   ├── eda                    # Notebooks for EDA
│   └── ml                     # Notebooks for ML Models
├── src                     # Source files
│   ├── utils                  # Utility Functions
│   └── test                   # Automated tests
├── reference_material      # Reference Materials
├── LICENSE                 # LICENSE File
├── requirements.txt        # Dependencies for CI/CD Workflow
├── Makefile                # Automated Script
├── CODE_OF_CONDUCT.md      # Code of Conduct File
├── CONTRIBUTING.md         # Contributing File
├── env.yml                 # Conda Environment File
└── README.md               # README

Usage

1. Clone the repository to your machine.
2. Set up the environment using the instructions below.
3. Template to include new API endpoint for training or prediction can be found in the data/raw folder.
4. To access the notebooks via Jupyter Lab, follow the instructions below.
5. To use the models directly, or to generate the reports, please refer to Makefile instructions below.

Environment set up

You can install all the dependencies you need using conda:

# Create and activate the environment
conda env create -f env.yml
conda activate api-risk-env

Please note that this process will take approximately 10-15 minutes.

Template for Input Data

The input data needs to be in a specified format for the script to run without errors. All the column headers are required to be present in the input data file. As for the data inputs, the script will accept empty fields. More specificially, all 18 column headers need to be present, however, not all 18 columns each row needs to be completed.

A sample of the data format, description of input columns and information about mandatory fields are provided in the raw_data_format.xlxs file inside data/raw folder.

Jupyter Notebook

To run the jupyter notebooks for analysis, please make sure you have install all the depenencies in the section Environment set up and run the following command

jupyter lab

Once the Jupyter Server is running, go to http://localhost:8888/ and navigate to the directory /notebooks.

Makefile

Makefile is a script that automates the process of training and testing the models with predefined parameters, in addition to generating the final report.

To run all scripts:

make all

To run preprocessing only:

make data/processed/preprocessed_train.xlsx

To create the model only:

make models/model.joblib

To predict only:

make data/processed/df_predicted.xlxs

To generate the final report only:

make book.html

Alternatively, you can run the scripts individually:

To run preprocessing:

python src/preprocessing.py --endpoint_path=<path_to_endpoint> --country_path=<path_to_country>--risk_rules_path=<path_to_risk_rules> --output_path=<path_to_output> --split_data=<bool>

To create the model:

python src/create_model.py --train_path=<path_to_train_file> --save_path=<path_to_save_file>

To predict:

python src/predict.py --model_path=<path_to_model> --predict_path=<path_to_predict_file> --save_path=<path_to_save>

Generate the proposal:

jupyter-book build docs/proposal_book/ --builder pdfhtml

Generate the final report:

jupyter-book build docs/report_book/ --builder pdfhtml

To build Github Pages, first:

jupyter-book build docs/report_book/

Then navigate to the docs/report_book/ folder and run the following command:

ghp-import -n -p -f _build/html

Contributing

License