It is hard to find a password that is absolutely secure / one that is not accessible to others. This app provides a personalized solution that is unique to only you through machine learning analysis of your own handwriting. Most importantly, you don't have to remember the password, as all you have to do is provide a handwriting sample

What it does

It provides an additional layer of security apart from general norms of password protection without the added hassle of memorizing another password -- this also protects against harassment from hackers and others trying to obtain or change your privacy settings (will not register of handwriting shows signs of undue stress)

How I built it

Found a database of training data -- Thoreau's handwritten journal. Sent these images through an online image converter (PDF -> JPG), cut them up into relevant sections, and fed them through IBM Watson's Visual Recognition API to train dataset to handwriting. Then used Android Studio to program frontend.

Challenges I ran into

Originally was working on a project that could analyze any piece of handwriting and tell you about the tester's underlying qualities -- however, training data was hard to find, so we pivoted. Additionally, IBM Watson's Visual Recognition API "was not available at this moment" after we had input training data. A few problems with configuring touchpad with Android Studio.

Accomplishments that I'm proud of

Came back strong from our original idea (pivoted in last 4 hours when we realized mathematics/implementation of our first idea was not feasible) -- and rallied into creating a prototype

What I learned

Learned a ton about graphology, was able to navigate and learn about API functionality, worked with an awesome team dynamic with new people (all met during the hackathon)

What's next for signature-analysis

Expanding functionality to analyze psychology with handwriting, integration with other password-protected apps, take digitalized notes from previous data (notes that you already have on Surface, etc) to use as additional training data

Share this project: