Vault Box

Framing the Problem

Traditional locks rely on keys or PINs, which can be lost, stolen, or guessed. They also offer limited real-time threat detection. We aimed to create a more intuitive and secure system by using behavioral authentication and introducing a dynamic risk metric based on sensor input.

Idea Explanation

We developed a knock-pattern recognition system that analyzes timing between knocks to authenticate users. Correct patterns grant access, while repeated failed attempts increase the system’s threat level. This allows the system to not only control access but also monitor suspicious behavior over time.

Implementation

Our microcontroller processes sensor data and knock patterns in real time, controlling a servo-based locking mechanism. This data is transmitted to a live dashboard, which displays system status, failed attempts, and threat levels. While currently connected via serial, the system is designed to extend to wireless communication and incorporate additional sensors.

Challenges

We had a very hard time to designing the physical box component and utilizing the servo, admittedly an error we are still struggling with. TIme has been a huge constraint and we were not able to get through implementing all of our features.

Accomplishments

We learned a lot about 3d design and CAD, as well as breadboarding as a whole. This being our first hackathon we're doing our best to view this as a learning experience. Although we were not able to complete all parts of our initial plan, we were able to work through core features and have a product idea that could be beneficial to millions.

Next Steps

What are the next steps for your project? How can you improve it?
To move VaultBox from a prototype to a fully functional system, we would focus on improving both hardware reliability and the intelligence of the decision engine.

On the hardware side, the current enclosure needs refinement. The servo mechanism needs to be properly fitted and mounted so the lock can physically actuate reliably. We would also expand the sensor suite beyond just tilt and shock to include additional inputs like sound or proximity to improve detection accuracy.

On the software side, the knock-pattern system can be made more robust by adding a proper pattern reset mechanism and improving timing tolerance. Right now, the system evaluates patterns, but it doesn’t fully handle edge cases or long sequences.

More importantly, we would enhance the decision engine to combine multiple signals instead of relying on single triggers. For example, instead of flagging an intrusion based only on tilt, the system could incorporate knock activity, timing patterns, and environmental data to make more accurate classifications.

Finally, we would move from a wired serial connection to a wireless system and improve the dashboard to support remote monitoring and alerts.