Dr. Doxter

Dr Doxtor's Portrait!

Inspiration

We were inspired to create this app because we ourselves are victims to self doxxing online. Sometimes the pictures we post would unintentionally reveal our personal information, so with this hackathon, we wanted to tackle this problem.

What it does

Dr. Doxter actively scans for images uploaded to discord servers and takes them down if they contain any sensitive information about the user. The user is then given the option to blur the image if they want to reupload it onto the server.

How we built it

Our project was built using python, through the utilization of libraries such as easyOCR, regex, requests, cv2, and discord.py. The base language we used to create our bot was python. We used discord.py to interact with the official Discord API. We also used requests to scrape images from the server. Then, we used easyOCR to extract text from images that are posted in the discord server. After, Regex helped us filter this data to determine what is “sensitive” or not. Lastly, Cv2 provided us with the tools to blur the sensitive information in the images.

Challenges we ran into

After we researched every tool we needed to create our bot, we spent hours testing out different algorithms to create our most efficient algorithm possible. We triumphed through and were able to piece each element of our program together to complete the puzzle that was Dr. Doxter.

Accomplishments that we're proud of

Our biggest accomplishment throughout the event was improving the efficiency of our detection algorithm. At first, Dr. Doxtor took 15-20 seconds to detect and take down sensitive image in servers. Now, it can do that in less than 5 seconds.

What we learned

This was our first time using the discord API. Discord is an app that is integral to our daily lives, so it was very cool for us to experiment and test with the backend of discord. This added a very unique skill to all of our arsenal's which we will continue to expand in the future.

What's next for Dr. Doxter

Here are the following improvements we can make to scale Dr. Doxter further 1) Detect languages other than English 2) Train our own AI model to identify sensitive data more efficiently and accurately 3) Branch out this concept to other social media platforms such as Facebook or Instagram