Anaheim, CA ― USENIX ― A security researcher suggested a practical solution on Wednesday for shielding users from the nuisance of robocalls. By employing an artificial intelligence assistant capable of assessing human-like interactions, individuals could effectively fortify themselves against these intrusive calls.
Addressing an audience of global researchers at the 32nd USENIX Security Symposium, Sharbani Pandit, holding a doctorate from the Georgia Institute of Technology, put forth a compelling proposition. She introduced an interactive virtual assistant called “RoboHalt” as a potential solution for users looking to fend off robocalls. This innovative assistant aims to counter calls from numbers that aren’t even covered by the existing call blocking applications’ blocklists.
Robocalls, a deceitful tactic employed by criminals, involve the use of prerecorded messages to deceive their targets. Those who are unfamiliar with such tactics, particularly the elderly and immigrants, are more vulnerable to falling victim. YouMail, a service focused on blocking robocalls, reported earlier this year that consumers in the United States were bombarded with over 50.3 billion robocalls throughout 2022.
Describing the functionality of RoboHalt, the researcher explained that the system operates as an initial point of contact for all incoming calls, engaging with the caller on the user’s behalf. This virtual assistant employs up to five questions, taking into account user convenience, to determine the nature of the caller—whether they’re human or not. If the system concludes that the caller is human, it will proceed to ring the user’s phone and forward the call. Conversely, if the model judges the caller to be non-human, the call is promptly blocked and rerouted to voicemail.
The researcher introduced seven distinct categories of questions that the virtual assistant relies on as its foundation to identify robocalls, including “Relevance” and “Speak up.” Under “Relevance,” questions such as “How are you doing?” and “How do you find the weather today?” are posed. On the other hand, “Speak up” involves instructing the caller to speak in a louder tone. While these queries might seem simple for human callers to address, they can create a real-time challenge for automated robotic callers to respond to convincingly.
“For example, at the beginning of the conversation, RoboHalt is more likely to ask a question that generally appears in the beginning of a phone conversation between two strangers,” Pandit explained to The Readable. She emphasized, “RoboHalt would never ask ‘Can you please repeat yourself?’ as a first question because they haven’t said anything.”
The researcher clarified that they undertook a security analysis to assess the virtual assistant’s performance. The results showed that RoboHalt successfully intercepted 95% of the 144 mass robocalls subjected to the evaluation. These mass robocalls are automated calls executed without the recipient’s awareness or consent.
Nonetheless, Pandit also underscored the inherent constraints of such a solution. She pointed out that thwarting scammers might pose challenges, especially when they leverage AI technologies to deceive others. Additionally, human fraudsters could potentially find it less difficult to engage with the virtual assistant, potentially circumventing its protective measures.
