This project is for CSE Machine Learning Challenge:
Challenge: Build a project that uses Machine Learning; Our criteria include application performance, focusing on speed, accuracy, model size, and resource allocation.
We were searching Kaggle when we were starting to lose hope on making something for this Hackaton. We stumbled upon one about discriminating between phishing links and regular links, we thought it would be an interesting challenge to tackle with our time. Even with the constant advertisements and warnings the Government of Canada outputs exposing these scams, we hear about the elderly, young kids, and unknowing adults falling prey to phishing schemes on the news daily. Taking a real life challenge and applying Machine Learning, and taking it to a point where people from any tech background could use it, seemed like a rewarding task-- One we couldn't pass up.
Link Ahead is a machine learning model and tool that can be used to discriminate between phishing links and non-phishing links with a 92.59% accuracy. Using this model, we constructed a Web Application to allow others to input any link of choice and test the trustworthiness of it. The background ranges from shades of red -- green to demonstrate the severity visibly. We also created a Google Chrome Extension that can be used to scan and test the safety of all the links available on any webpage.
To build this tool, we used a dense sequential model trained on over 1.4 million URLs via TenserFlow. To build the web app + API, we used Flask, HTML/CSS/JS and data processed via python from the Kaggle source. To build the Google Chrome Extension, we used HTML/CS/JS, while calling to the sketch factor API repeatedly.
- Getting a Flask application to be hosted on Vercel [UnSolved - Went with 3 other options then decided to set up out own server]
- Editing Chrome's Manifest to allow access to all web pages [Solved]
Despite it all, I was able to develop a product that is usable and a helpful tool for everyone’s day to day life. Big thank you to @jar who walked me through every major setback that occured along the way.
how to use Flask to build a webpage and API. how to utilise the pandas library dataframes to commute row calculations instantly. how to call APIs within a Chrome Extension.
Find somewhere to host indefinitely, and upload the chrome extension for approval to be posted on the Chrome Web Store.
[Github] https://github.com/oZep/uOttaHack6-Link-Ahead
[Slides] https://docs.google.com/presentation/d/1_gk9gpgYydGDOnoFh71wkYx1LNCpRqqVUksM8woCePU/edit?usp=sharing
[Devpost] https://lnkd.in/eTaZUyCx