Inspiration
Every year, there's always that one story of a family loosing their life-savings to the classic phone scam. Inspired by Google Pixel's Screen my Call feature, our solution aims to protect those most vulnerable by leveraging the latest Machine Learning, Generative AI and Text-to-Speech Models.
What it does
Scam-Mah is a two-part comprehensive solution designed to be integrated into the calling experience.
Machine Learning Algorithms identify anomalies within the cellular network and flags callers with atypically-high call rates and atypically-low average call durations as potential spam. Integration of this algorithm within a telecommunication company will allow for real-time intervention to warn recipients of potential scam.
When an incoming call is suspected as a scam, users are given the option to allow AI to accept and handle the incoming call. Generative AI and Text-to-Speech Models are used to process the incoming call and deliver a humourous response intended to waste the scammer's time.
The goal of the project is to provide a light-hearted and humourous solution to the everyday phone scam. By wasting the scammer's time, they will not be able to scam as many other people and would-be victims are protected. The humourous twist naturally encourages consumers to leverage the tool, while turning a potential scam into reverse spam.
How we built it
StandardScaler and IsolationForest ML algorithms power the anomalous detection model, while Google's Gemini and ElevenLabs's TTS APIs power the live response component of Scam-Mah. The backend is handled by Python and integrated with an HTML/CSS/JS frontend.
Challenges we ran into
Implementing the integration and communication between the Python backend and the JavaScript frontend proved to be much more difficult than anticipated. What we thought would have been a simple function call turned out to be a 3-hour struggle of implementing the Flask library, reorganizing the file hierarchy, and learning to create post and fetch commands.
Accomplishments that we're proud of
We are glad to have been able to learn so much and deliver a functional prototype within just a 24-hour hacking period.
What we learned
Facilitating API calls within Python; Integrating Python backend with HTML/CSS/JS frontend through Post and Fetch commands; and Implementing libraries within Python.
What's next for Scam-Mah
Scam-Mah is ultimately designed to be integrated seamlessly into the calling experience, where ML and spam detection is handled on the telecommunications end and the live response can be invoked natively from the call screen. In the interest of time, a proof of concept was implemented via a demo website; the next step would be to develop Android and IOS versions.
Log in or sign up for Devpost to join the conversation.