DocuShield

Unfiltered Document
Filtered Document

Inspiration

We were discussing how easy it is to accidentally send documents with sensitive information onto the internet and wanted an app that would censor all private information in a document with the click of a button!

What it does

DocuShield takes a Microsoft Office word document as input and with Natural Language Processing it detects possibly sensitive information and removes it. Then, the app exports a copy of the censored document.

How we built it

We used Python as our coding language and GitHub for version control. Originally, our program was using ReGex filters to find patterns in text, but we found this to be limiting. A more dynamic approach was to use the Natural Language Processing of the Spacy Python library. Then, used the tkinter library to create a simple GUI, and docx to manipulate Microsoft Office documents.

Challenges we ran into

Our biggest challenge was getting the different Python libraries we utilized to install and work properly on each of our laptops. This became problematic, as we were forced to change the versions of Python and it lead to a lot of headache in the early hours of the morning. Besides that, our second biggest challenge was to find a dynamic and flexible way to find sensitive information in a document without explicitly coding patterns to find.

Accomplishments that we're proud of

Starting the project by using "hard-coded" Regex filters, and ending the program with an AI that covers more patterns in documents than we could've ever imagined.
Building a solid GUI for the user to navigate easily.
Being able to break the project into separate portions and multitask by using GitHub