🔖 Fast forward to the predictive models

About

Sanitara is a machine-learning and heuristics based Parental Control System, designed to give the user a prediction about a given domain names' child-friendliness.

This is a PoC developed using the RAD (Rapid Application Development) method with the core models and functionality being completed in less than 12 hours.

How it Works

Currently, the service works by exposing an API, in which one can send a domain name, and recieve a safety prediction via. a series of heuristics and machine learning models. At this time, predictions for both phishing and pornography websites are complete and fully working as well as white- and blacklist functionality.

To see detailed metrics and in-depth explanations of how the models are built check out models.md. And for an explanation of which heuristics we use go to heuristics.md

Running the Service

⚠️ these are predictions, not facts: not every result will be correct.

$ git clone https://github.com/frederikgram/sanitara.git
$ cd sanitara
$ python -m venv venv
$ ./env/Scripts/activate
(venv) pip install -r requirements.tx
(venv) python start_service.py

Testing the Installation

$ curl http://127.0.0.1:8000/api/v1/predict/?message=https://www.badwebsite.com
>>>
{
  "predictions": [
    {
      "name": "pornography",
      "prediction": 0,     # is not pornography
      "err": null
    },
    {
      "name": "phishing",
      "prediction": 1,     # is a phishing website
      "err": null
    },
    {
      "name": "whitelist",
      "prediction": 0,     # was not found in whitelist
      "err": null
    },
    {
      "name": "blacklist",
      "prediction": 1,     # was found in blacklist
      "err": null
    },
    {
    "name": "regex",
    "prediction": 0,       # did not match any given regex pattern
    "err": null
    },
    {
    "name": "keyword",
    "prediction": 0,       # did not match any given keywords
    "err": null
    }
  ],
  "err": null,
  "input": "https://www.badwebsite.com"
}

FastAPI

FastAPI is a framework designed to serve production quality APIs' without to much effort or overhead. We're using this to deploy our machine learning models, and automatically generate documentation for our API.

Documentation

After running the service, as mentioned here. It's as simple as navigating to http://127.0.0.1:8000/docs to see the documentation as provided by swagger