Inspiration

The need for an internal, feasible, and user-friendly authentication method. Also, one that is difficult to brute force.

What it does

Using the webgazer library, a user calibrates their individual gaze to match points on the screen using the webcam. We divided the screen into 9 separate sectors. The "secret code" or way to authenticate is a sequence of looking directly at the sectors on screen, like a passcode. The gazer will record and map one's gaze and assign it to one of the 9 sectors. If the sequence of viewing is correct, access will be granted.

How we built it

Using the existing webgazer library, we first defined the coordinates of the 9 sections on the webpage. Then, we implemented a classifier that, in real time, takes in and classifies the current gaze coordinates on the screen to one of the 9 sections. If the gaze has been in one section continuously for 150 milliseconds, it will be added to the list of inputs (this is to ensure the user was actually looking at the correct section). Once 3 sections have been added, the program compares the inputted pattern to the actual pattern of views needed to unlock the machine. If it is correct, the machine is "unlocked" (displays success message). Else, the user gets 4 more tries to unlock (2 for demo purposes).

Challenges we ran into

· Understanding the open source webgazer library to use its functionalities in our project. · Defining and dividing the sectors within the webpage in terms of size and margins in between them. · Training the classifier correctly in order to minimize the prediction stutter which prevent identifying sectors correctly.

Accomplishments that we're proud of

It works! We were able to get the demo up and running, and anyone should be able to try it if they like. Just having it all come together in such a short time is something that we're really proud of.

What we learned

Building one step at a time: hardcoding the answer to a problem, then making it dynamic can be very helpful. Being patient also helps, and looking at examples. Good and clear communication among teammates is also very important.

What's next for Eyeconic

Being able to dynamically put in a custom pattern with your eyes, then unlock with that pattern. We would also have to reskin the alerts, and somehow locally encrypt the password for comparison. Blink detection and facial recognition are also avenues to be explored.

Built With

Share this project:

Updates