Gama (2^nd runner up in Locus codecamp 2023)

Our project revolves around the idea of building a browser extension that empowers users to navigate the web with confidence by detecting and warning them about phishing sites, fake reviews, and other online threats.

Requirements

• Python 3.11
• Whois cli
• pip

Installation

Install gama with git & pip

    > git clone https://github.com/PhuyalGaurav/gama.git
    # now be in the gama folder & make a venv
    > cd gama
    \gama> python -m venv venv
    # install all the dependencies
    \gama> venv\Scripts\activate
    (venv) .. \gama> pip install -r requirements.txt

Deployment

To deploy this project run

  # be in the root gama folder
  # make sure you have your venv activated
  \gama>venv\Scripts\activate
  (venv) .. \gama> cd backend
  (venv) .. \gama\backend> python app.python
  # Now the local server should be running

Use of some thechnologies and why?

Libraries Used	Why?
Fast API	We Used to fast api. Because it is very light weight and also new!, we also went with a api approach so that other developers can use our api to integrate it in thier own applications
Sk Learn	Since our data set was quite small . We went with sklearn rather than tensorflow. Also with our limited hardware we could train with it easier than tensorflow that utilizes heavy & expensive cuda cores
Whois	We choose whois cli instead of an api because we wanted to keep this as free as possible and also keep the delay between front & backend small.
Chrome extention	we went with chrome extension to make this as usable as possbile & reach as many people as possible we are planning to add firefox in the future so that it reaches mobie uses too

File tree

gama/
├─ .github/
│  ├─ 0.png
│  ├─ 10.png
│  ├─ 11.png
│  ├─ 12.png
│  ├─ 8.png
│  ├─ 9.png
│  ├─ image-1.png
│  ├─ image-2.png
│  ├─ image-3.png
│  ├─ image-4.png
│  ├─ image-5.png
│  ├─ image-6.png
│  └─ image-7.png
├─ .gitignore
├─ LICENSE
├─ README.md
├─ backend/
│  ├─ Datasets/
│  │  ├─ amazon_reviews_2019.csv
│  │  ├─ fake_news_dataset.csv
│  │  └─ phishing_site_urls.csv
│  ├─ MLModels/
│  │  ├─ fakeNewsModel.pkl
│  │  ├─ fakeNewsVectorizer.pkl
│  │  ├─ phishing.pkl
│  │  ├─ reviewModel.pkl
│  │  └─ reviewVecotorizer.pkl
│  ├─ app.py
│  ├─ db/
│  │  └─ db.sqlite3
│  ├─ helpers.py
│  ├─ models.py
│  └─ news_predictor.py
├─ frontend/
│  ├─ background.js
│  ├─ contentScript.js
│  ├─ logo.png
│  ├─ manifest.json
│  ├─ override.html
│  ├─ overrideJs.js
│  ├─ overrideStyles.css
│  ├─ popup.html
│  ├─ popup.js
│  └─ styles.css
├─ proposal.pdf
└─ requirements.txt

• .github/ contains all of the images needed for github read me
• backend/ contains the backend code
  • datasets/ contains all data sets used to train our models
  • MLModels/ contains all the machine learning models
• Fronend/
  • contains code for extention

Python files

app.py

Contains all the code for running the fast api server.

helpers.py

Contains all the helper function & Review model with its text pre processor

news_predictor.py

Contains the class for preding if a news is Fake or not.

models.py

Contains database models and pydantic shecma models.

API Reference

Check For phishing

  GET /phishing?url=

Parameter	Type	Description
url Required	string	Gives true if phishing link detected else false

Get all reports

  GET /reports

Parameter	Type	Description
None	none	Returns all the reports that are repoted by user (saved on db).

Get details about a website

  GET /details?url=

Parameter	Type	Description
url	string	the link of the website whose details is neededs

value returned	Type	Description
Name	string	Returns all the reports that are repoted by user (saved on db).
registrar	string	The registrar of the domain.
registrant_country	string	The country of the registrant.
creation_date	Date	The date when the domain was created.
expiration_date	Date	The date when the domain will expire.
last_updated	Date	The date when the domain was last updated.
dnssec	bool	The registrant of the domain.
registrant	string	Returns all the reports that are repoted by user (saved on db).
emails	string	The associated emails of the domain.
country_name	string	The country name of the domain.

Report a website by a user

  POST /report

Parameter	Type	Description
Url	string	url of the website which is to be blocked
reason	string	the reason why the website should be blocked

Report a mis identified website by a user

  POST /report_mistake

Parameter	Type	Description
Url	string	url of the website which is to be unblocked
reason	string	the reason why the website should be unblocked

Check if a review is real of fake through ML

  POST /review

Parameter	Type	Description
review	string	The review which is to be checked

Value returned	Type	Description
prediction	bool	True if review is fake

Check if a news is real of fake through ML

  POST /news

Parameter	Type	Description
news	string	The news which is to be checked

Value returned	Type	Description
prediction	bool	True if review is fake

Make a specic review to primary check

  PUT /reports/{id}?real=

Parameter	Type	Description
id	integer	the id of the report to be made ture or false
real	bool	The boolen value of the report to be set.

How to use the extention

• After the server has been set up. Add the frontend extention to your web browser . Make sure that developer mode is enabled.

• Now Load unpacked

• Now select the frontend folder from gama directory

• Now the extention should be installed. go ahead and pin it. so:

• Now the extention is installed!

• Your extention should be shown like this with details

• Now whenever you visit a suspisous site the above screen is shown protecting you from online threats

• If you wish to Disable this just click the checkbox

• If you ever find a site that seems suspicous you can report it by just reporting it in the textbox like so:

• If you want to check for a fake review then highlight the text and right click

• Now click on gama and click on Review check

• Now you should get a notification like this:

• If you would like to check for fake news follow the same process just click on news check :
• Then you should get a notification like this

Appendix

• The python server is easily uploadable to the web & serve use through online apis
• The apis can be used by 3rd party apps too.
• The machine learning models are trained on datasets found on kaggle (With the creators permission )

Model accuracies (pre tested)

• Fake News Model : 81.23%
• Phishing Detector Model : 78.93%
• Fake Review Model : 73.22%

License

MIT

Refrences

Thank you manideep2510 for letting us use their pre-trained fake news detection model

His repo

Authors

• @phuyalgaurav90

• @munannoo