Scam Stopper

Inspiration

Senior Citizens lose money from fraudulent phone calls frequently. And while federal officials say Millennials are twice as likely to be scammed as seniors, they’re less likely to lose as much: Consumers in their 70s lost an average of $621, while victims ages 80 and older lost nearly $1,100. Millennials, in contrast, lost $400 on average, according to the FTC.

What it does

Scam Stopper analyzes phone calls as they are happening from an unknown number and listens for any warning signs to indicate that the call is a scam. Some common fradulent calls are from people who claim to be from the IRS, Social Security Administration or Banks and ask for credit card information, social security numbers, and payment information.

How we built it

We used Azure Microsoft Speech to Text Analysis to turn the phone call into text. We then used Azure Keyword Extractor to analyze the common keywords and compared them to our word bank of trigger words to see if any of them matched. We also used Microsoft Sentiment Analysis to analyze the text to detect if the text was threatening or overly excited, any extremes were taken into account when detecting if the call was a scam. If there were enough keywords to trigger our thresholds we would notify the bank to watch out for any transactions in the next 15 minutes and to keep a close eye. We used flask and jinja as our web framework, and used jQuery in our web pages.

Challenges we ran into

We had an extremely hard time displaying text asynchronously as we spoke into the microphone to display on Flask. We tried threading, callbacks in flask,jquery post requests, but it just wasn't able to display text as we spoke onto our webapp. Connecting the frontend to the backend was extremely difficult.