TOSbot

Inspiration

After reading about the vast amount scandals occurring over the handling of user data by large corporations, I began to ponder why people, myself included, don't read the terms of services before providing these companies with their personal and sensitive information. Then I actually took a look at one and reminded myself why. There are two main reasons why these service agreements are a pain to read: the length and the convoluted language. Given my experience with building machine learning algorithms, I decided to put my skills to the test.

What it does

TOSbot is a web extension which scrapes the privacy policy of the websites and services that users use and analyzes them, breaking down the good, bad, and ugly parts of the agreement while ranking how privacy-friendly that service is.

How I built it

I built the frontend of the project using python, specifically flask, and javascript to manage the extension. The web scraping implementation, along with the final result html page was also part of the frontend. The backend was built using a neural network written in python. The neural network was created with the keras library which runs on a tensorflow background. The Neural Network was trained to find and identify important parts of the service's privacy policy.

Challenges I ran into

I ran into a multitude of issues on the frontend, including communicating between javascript and flask and effectively and accurately scraping the key parts of the different websites privacy policy. That was plenty frustrating but luckily the backend was in a good mood and I was able to effectively build, train, and test the NN with considerably less effort. The biggest challenge by far, however, is the lack of a partner to help problem solve with me. That would have made life much easier

Accomplishments that I'm proud of

I'm proud of the fact that I got the backend and frontend to cooperate with each other and the fact that I actually succeeded in building TOSbot.

What I learned

I mainly learned how to be patient and keep my cool when things were not working. I also learned how to build a chrome extension and how to make that run a NN in the background.

What's next for TOSbot

I am hoping to pretty up the frontend in the future and make the NN perform more accurately.

Built With

• flask
• html
• javascript
• keras
• logic
• patience
• python
• scrapy
• tensorflow