Inspiration
Cyber Security is becoming a growing need in each person’s life as more and more people get connected online. Since the pandemic hit in 2020, the crypto theft rate skyrocketed by 516%, which made scammers billions of dollars. This is what inspired us to do this project.
What it does
Safe Crypto is a one-stop-shop where you can verify your transactions to see if they are ransomware free, get correct filtered information/News on Crypto (Especially Bitcoin), get recommended sources to learn how to trade, and lastly, get recommended the best in class brokers in the markets to trade your cryptocurrency. The user inputs the bitcoin address the bitcoins were sent to, the date of the transaction, and the amount sent. The model takes these inputs and predicts if the transaction was ransomware or not. The icons on the sidebar are what we call the "Information & Resources Field", where the user can get the latest information and resources on bitcoin.
How we built it
We broke up our build process into 3 different categories: Design, Machine Learning, and Website integration.
-> Design: Bitcoins are the new age currency and we understand the need to attract new audiences to this technology. Hence, we came up with a theme inspired by the same. We started off by trying to understand the needs of the target audience using a user persona, we also did a trend study to keep up and come up with best-fitting design ideas and explored a variety of visuals and imagery that are inspired by the theme. It's important to note that we designed the whole website from scratch and this is a completely original design
-> Machine Learning: We built & trained our ML model from scratch and used Supervised Learning Models to train our model. We used our own data sources (~3 million Entries!), taken from the UCI database. After our analysis, we found that Random Forests do a really good job generalizing with the data. We Hyperparameter tuned the model to get the best parameters as Max_depth = 10 and n_estimators = 30. Our trained model acquired a 99.8% accuracy in detecting potential ransomware.
-> Website Integration: We used react javascript and bootstrap, connected it to a backend server using flask. We used forms to get the user to input various fields for the result to show up. We used buttons to verify the user data.
Challenges we ran into
We ran into a lot of challenges connecting the front end to the backend. Connecting the user forms to output a certain response from the backend was a little glitchy. Coming up with a unique and minimalistic design was really difficult as the design process itself is an iterative process. In terms of machine learning, some of the variables got interchanged which lead to the model outputting wrong results initially.
Accomplishments that we're proud of
We are proud of our design, accuracy of our Machine learning model, and getting it to work on the website
What we learned
We learned to work together as a team, distribute work, communicate with each other, and most of all, how to get through difficult stages
What's next for Safe Crypto
The only thing we weren't able to do was to implement the complete design that we had in mind. As we had issues changing the background in the code last moment, we had to ditch it.
Log in or sign up for Devpost to join the conversation.