Inspiration
AI safety is a very privatized tool and not as easily accessible to the public. We are trying to commercialize AI security to reduce the stigma around AI
What it does
An AI authentication tool to protect governments, celebrities and users from malicious AI based attacks that involve the spreading of misinformation.
How we built it
We built it using TensorFlow and Python along with a number of libraries, including numpy, pandas and keras for the backend processing and React, JavaScript, and Tailwind.css primarily for the frontend.
Challenges we ran into
One of the challenges we faced on the backend side was being able to bring in and process large amounts of image data. The kaggle dataset proved to be larger than our systems could handle at the time so we ended up parsing in a portion of the entire set. We also encountered quite a few errors while identifying the different parameters we could use for processing the images. Routing became a huge problem in regards to navigating .jsx files.
Accomplishments that we're proud of
We came into this hackathon with very little AI and ML experience and not much experience doing backend either but we were very proud of being able to learn quick and be able to construct an ML algorithm.
What we learned
How to create and work with frontends as well as develop complex machine learning algorithms. In addition, learning new concepts in areas that we saw our teammates needed some assistance on.
What's next for Auth Insurance
We plan to continue to develop the Machine Learning Algorithm to detect deepfakes at a higher accuracy. Furthermore we would like to be able to detect not just images, but audio and video deepfakes as well. A web scraping tool will also undergo development which will be able to constantly scrape popular social media sites (Instagram, Twitter, Meta) as well as google in order to search and detect the moment these are posted. With these social medias, we hope to bring consistency and efficiency to increase awareness and identify threats as fast as possible. Improvements on the notification system on how we warn our clients when a something of this nature is posted will also be on our agenda.
Log in or sign up for Devpost to join the conversation.