A photo from Jupyter Notebook, from initial design phase
A photo from our Streamlit web app

SwipeShield Project Story

Inspiration

The idea for SwipeShield was born out of a growing concern for mobile security. With the rise in smartphone theft and unauthorized access, we wanted to create a solution that improves user authentication while maintaining convenience. The concept of using unique swipe patterns as a biometric measure intrigued us, leading us to explore how machine learning could be used to analyze these patterns effectively.

What it does

SwipeShield is a user verification system that analyzes swipe gestures on mobile devices to distinguish between legitimate users and potential imposters. By employing machine learning algorithms, our app identifies swipe patterns and provides real-time feedback on whether the last few swipes match the user's behavior, improving security without compromising user experience.

How we built it

We began by gathering swipe data using the Android Debug Bridge (ADB) to collect the necessary input for our machine learning model. After cleaning and organizing the data, we employed K-Means clustering to analyze swipe patterns and classify them into user and non-user categories. The project was developed using Python for the backend and Streamlit for the interactive web application, allowing us to visualize data and results in a user-friendly manner.

Challenges we ran into

One of the primary challenges was ensuring the accuracy of our machine learning model. Tuning the parameters for K-Means clustering took several iterations, as we needed to balance the sensitivity of the model with its ability to avoid false positives. Additionally, integrating the data collection from ADB into a seamless workflow proved difficult, requiring us to troubleshoot various technical issues along the way.

Accomplishments that we're proud of

We successfully built a functional prototype that demonstrates the potential of using swipe patterns for user verification. Our interactive web application showcases real-time analysis of swipe data, allowing users to visualize their swipe patterns and understand the clustering results. Furthermore, we are proud of our ability to work collaboratively as a team, leveraging each member's strengths to overcome obstacles.

What we learned

Throughout this project, we gained valuable insights into the process of collecting and analyzing biometric data. We deepened our understanding of machine learning algorithms, particularly in the context of unsupervised learning. Additionally, we learned the importance of user-centered design, ensuring that our application is both secure and user-friendly.

What's next for SwipeShield

Looking ahead, we plan to refine our model by incorporating more diverse datasets to improve its accuracy and reliability. We also aim to explore additional biometric authentication methods, such as touch pressure sensitivity and swipe speed, to improve security further. Our goal is to make SwipeShield a comprehensive security solution that adapts to users' unique behaviors while remaining effortless to use.