Sharbani Pandit - Fighting Voice Spam with a Virtual Assistant
From Kathryn Gentilello on February 24th, 2020
Telephony has been a trusted channel in the past but technological advances have exposed it to abuse that is perpetrated by fraudsters and criminals. Mass robocalls, call source spoofing and voice phish- ing are some of the abuse techniques that are being used to target millions of people. Because of the serious nature of the problem of unwanted spoofed robocalls, there are several ongoing initiatives that are aimed at combating caller ID spoofing. However, quick and at scale deployment of such solutions is unlikely. We address the problem of robocalls and other unwanted calls by introducing a virtual assistant that can be integrated in smartphone applications. Similar to a human assistant, the virtual assistant can pick up an incoming call and use audio analysis techniques to screen it to determine if the call is unwanted. By developing a proof-of- concept prototype of the virtual assistant, we demonstrate that a small amount of audio at the start of a call can be used to make such a determination. A user study conducted by us showed that our system, RobocallGuard, is able to preserve the call experience of a phone call and does not negatively impact legitimate callers. We believe the virtual assistant can help overcome the limitations of existing defenses such as phone blacklists, which can be easily undermined by caller ID spoo ng. Since blacklists are the basis of several commercial solutions, our approach could o er the next line of defense against telephony abuse.