Spoofed calls with AI voice impersonation is a surefire way to scam anyone of their life savings. To combat this, caller ID verification and call monitoring is needed.
We detect likely spoofed calls using the STIR/SHAKEN protocol. If a caller cannot be verified to actually be who they say they are, we monitor the call using an AI Edge Model, DistilBERT, that has been fine tuned on an anonymized scam call dataset (from "Classifying Scam Calls through Content Analysis with Dynamic Sparsity Top-k Attention Regularization"). This allows us to give a notification when a call has a high likelihood of being a spoofing scam.
This notification allows the user to be more wary of the caller, and to redial the alleged caller from their contacts to bypass the spoofing. In the case of a false positive, the only impact is that the user will immediately redial the caller and continue the conversation.
Built With
- kotlin
Log in or sign up for Devpost to join the conversation.