Inspiration
The rapid increase in phishing attacks and online scams, especially during the COVID-19 pandemic, highlighted the need for more proactive and real-time protective measures.
What it does
The browser extension scans page urls, qr codes and content of emails and sms to find out if the content inside is safe for browsing. Using the traffic light system, it shows the following colours;: Green for safe, Yellow for caution as it may have unknown factors that cannot determine its safety, or red to indicate that the browser is not safe, while the Web App can do functionally the same, without the indicator of traffic lights.
How we built it
We built it with a flask backend with vanilla js. It uses sqlite as its database, and uses scikit-learn to train the ai models to detect phishing with NLPP We used pandas to help process and clean our dataset as well.
Challenges we ran into
It was hard to get the browser extension working as it was our first time making one. It was difficult to find a good dataset so that our ai can train on accurate data or accurate results.
Accomplishments that we're proud of
We are proud that we were able to complete this project in such a short period of time. Besides just coding, we had to brainstorm ideas, design the UI, design the database, coming up with the features of the app. We are proud that we were able to finish the ai model which reaches up to 98% accuracy on the.
What we learned
We learnt about AI and ML algorithms as well as a bit of data science as we had to make the phishing detector. We also learnt how to make browser extensions which are useful as they can run background scripts when user is browsing as usual.
What's next for PhishNet
We can look into improving our AI even more with a larger dataset, better algorithms like Neural Networks and the new transformer based models. We can also use pretrained models like bert and gpt to further increase accuracy. For the browser extension, we add in AI in the future to detect malicious traffic that may have passed the security checks. We can also make the extension remember traffic before the current one so that it can correlate traffic between them and find malicious traffic more accurately.
