Recently, one of our team members, Trevor, got his credit card information stolen and we realized how much of an issue scam calls are in today's world. We want nobody to go through the same experience. So we created ScamSlam!

What it does

ScamSlam is a mobile app that you can sign up for which gives you a new phone number. This phone number is then linked to the user's actual phone number. This allows for the calls to be forwarded to the user after being scanned. When an incoming call is received, ScamSlam streams the call data to intelligent machine learning models which transcribe and predict in real time the likelihood the call is a scam. If a scam is detected ScamSlam SLAMs the call to protect the end user. After the SLAM the end user is notified that a scam was detected and the call was ended.

How we built it

Twilio was used to generate a unique phone number for each user and intercept the call. Node.js, Google app engine and Google Cloud Speech-to-Text api were used to get the phone call data of the hacker. IBM Watson machine learning classifier was used to train on the spam/not spam data set of emails. React native was used to build the iOS app.

Challenges we ran into

It was quite a challenge to integrate the machine learning model with the active phone call to get real time results. It was also challenging to make the phone call disconnect on its own once the call was classified as spam.

Accomplishments that we're proud of

We were able resolve our challenges and build a fully functioning app! It was fun to integrate the backend and the frontend together.

What we learned

How to stream and process data in real time, deploying to Google Cloud, how to train machine learning models in the cloud.

What's next for ScamSlam

Integrate ScamSlam technology with Twilio Elastic SIP trunking to allow anyone to port their existing number into a VOIP network and be protected from telephone scams. Additionally, continue to improve our scam detection models.

Share this project: