Questions for ‘A new tool could guard against deepfake voice scams’ 

an illustration of a robot hand holding a telephone reciever with a soundwave in the shape of a person coming out of the speaker

Scammers can use AI to create deepfake mimics of people’s voices. But researchers are fighting back. They’ve found a way to trick the tricksters and protect others from impersonating your voice.

Moor Studio/DigitalVision Vectors/Getty Images Plus

To accompany A new tool could guard against deepfake voice scams’  


Before Reading:

  1. Websites and phone apps use various techniques to confirm your identity. Besides password protection, list two identity authentication techniques. Which of these techniques do you feel is most vulnerable to attack by identity thieves? Describe one major weakness of this technique.
  2. If a thief were to steal your identity, which two websites or apps would you be most concerned about the thief accessing? For one of these, explain how such a theft might affect you. Briefly describe the protections this website or app uses to confirm your identity.

During Reading:

  1. Which famous YouTuber’s likeness appeared in the fraudulent October TikTok clip described in this story? What did fans believe they were purchasing when they paid shipping fees?
  2. How does AntiFake’s approach differ from most deepfake defenses?
  3. When identifying a voice, people pay more attention to some frequencies of sound waves and less attention to others. How do deepfake AI models differ from us in how they analyze sound?
  4. Does adding AntiFake protection to an audio clip increase, decrease or have no effect on the clip’s sound quality?
  5. How many deepfake AI models did Zhiyuan Yu and his team use to test AntiFake?
  6. After adding the AntiFake protection, what percentage of the 600 deepfake voice clips used in this study tricked people and machines?
  7. Besides adding background noise, how might future versions of AntiFake modify someone’s voice?

After Reading:

  1. Do you think the deepfake protection described in this story is valuable in its current form? Answer by rating this technology’s value on a scale of 1 to 10, with 10 being the highest. Explain your rating by describing either (a) this technology’s greatest strength or (b) this technology’s greatest weakness.
  2. Briefly explain how AntiFake alters a voice clip in such a way as to fool deepfake AI models but does not fool human listeners. How might deepfake AI models overcome such protections in the future? Do you think voice authentication will be a secure way to confirm identity in five years? Explain why or why not.