During my experience I have always managed to forget what my password was. That's why we built Sign Signature: the same signature that you have learned over the years can be used as an authentication system to unlock access to various native and web services.

What it does

Sign Signature can be used as either an alternate login system or as a part of a two-factor authentication system for even more security. After logging in with their username, users will be prompted to sign their signature on our iOS app, which acts as a medium of collecting signatures, allowing the user to then gain access to a variety of services which use our authentication system. The signatures are then verified via a convolutional neural network, which has been trained to identify different signatures corresponding to each user. Once the server verifies the signature is a match, the user can login is safely.

We implemented our authentication system on both web and native applications, showing its versatility. The native application implements two-factor authentication, requiring a password as well as signature verification. On the other hand, the web application is lightweight and only requires a quick signature to log in. Once signed in securely, the user can view their dashboard, which has services such as announcements and control of a wireless Ingenu unit.

How we built it

The main verification system in Sign Signature is a convolutional neural network. Instead of simply using an API, we wrote and trained a network ourselves using Tensorflow. This allowed the network to be customized to fit each data pair as closely as possible.

The iOS and OSX applications were written in native Objective-C, while the web app was created in javascript.

The two servers handling verification are written in Python, and hosted on a Microsoft Azure instance. These servers also serve as an access point to an Ingenu unit.

Challenges we ran into

Transferring image data from the iOS into the neural network caused a lot of issues, as converting into base64 for transfer originally corrupted the image due to newline characters in the string.

Hardware we were planning to use was not available and we had to pivot our idea.

Accomplishments that we're proud of

it worked low key weeb background

What we learned

how to make it work

What's next for 251 Sine Signature

make it work better

Share this project: