FlashGuard

HACKWPI 2019 SUBMISSION
By Bryce Corbitt, Brian Earl, and Jeffrey Harnois

Directories and files included in this project:

FlashGuard
|-- LICENSE
|-- README.md
|-- browser
|   |-- browser.py
|   `-- client.py
|-- config.json
|-- main.py
 -- server
    |-- extension_handler.py
    |-- video_process.py
    `-- wait_for_page_change.py

How it works

The Chrome window that opens upon running this project is monitored using Selenium, a browser automation platform. When the user visits a youtube video url, Selenium detects it and sends a TCP request to a server side which we have implemented. The server side downloads the youtube video and parses frame-by-frame to detect flashes within the video. If enough flashes are detected in the video in a short enough time, the server will respond with a message that makes Selenium pause the video and create a warning pop-up. Additonally, videos that have already been processed have their results stored persistently so that recurring videos are handled much faster by the server should they be viewed again.

Installation

• Install dependencies
  pip3 install opencv-python youtube_dl numpy selenium
• Download chromedriver and configure chromedriver path in config.json
• Configure hostname of server in config.json
• Start server:
  python3 server/extension_handler.py
• Start client:
  python3 main.py

Pop-Ups

After a video is processed, it wil respond with a reccomendation to whether the video could be potentially harmful
or not. Videos that are processed as harmful will be paused and produce give pop-ups that look like this:

This project currently only supports videos that are 30fps. If a video doesn't get processed because of this, it will be paused and given this alert as a warning: