It only takes 15 seconds of audio to clone a voice
Artificial intelligence is becoming more advanced. While still in its infancy, its text-to-speech generation tool is capable of generating a simulated voice using a short 15-second sample of an individual’s voice and has recently become more readily available. Other tools need only about 3 seconds of audio to create a compelling fake. People are already taking advantage of this technology by creating AI generated spam calls of famous people. But the scams can go much further than that, and they can have costly consequences.
How its done
- Data Collection: The first step involves collecting audio data from the target individual whose voice is to be cloned. This could include recordings of phone calls, interviews, speeches, or any other publicly available audio. For a borough or municipality, this could include recorded meetings that are then placed online.
- Training the AI Model: The collected audio data is then used to train a deep learning model, typically a variant of recurrent neural networks (RNNs) or convolutional neural networks (CNNs). These models learn to understand and replicate the nuances, intonations, and speech patterns of the target voice.
- Voice Synthesis: Once the model is trained, it can synthesize new audio clips in the cloned voice. Users can input text, and the AI model will generate speech output in the cloned voice.
- Fine-tuning and Improvement: The synthesized voice may initially lack clarity or authenticity, so scammers may employ techniques to fine-tune the model further. This could involve additional training on specific phrases or contexts to improve the quality of the synthesized voice.
- Deployment: After the voice cloning process is complete, scammers can use the synthesized voice for various fraudulent activities, such as impersonating the target individual in phone calls, creating fake audio messages, or manipulating audio in videos.
The Rise in Voice Scams
There are your run-of-the-mill scams like impersonation scams where scammers pretend to be someone you know or trust to gain access to PII or accounts, fake evidence scams where they try to falsely incriminate individuals, or voice-based ransomware where scammers threaten to release manipulated audio recordings. None of these are great, but where it gets even scarier is where your bank account (or your municipality’s bank account) is concerned.
Scammers can compile enough of your voice to get the bank to transfer funds over the phone.
A lot of banks use voice biometrics as a security measure. However, the security risk is less here as there are usually more layers of security involved. When getting through to a bank employee, scammers can usually social engineer the situation, convincing the person on the other end to do their bidding. By creating a high-pressure situation with the bank employee, scammers can get their way. It isn’t necessarily the scammer talking to the employee using the scammers voice. The scammers are likely using text to speech using the AI generated model of your voice.
Where are our voices online?
It is relatively easy to find our voices on the internet these days. Even if you’re not a public official or YouTube star, you probably have a social media account. Snippets of audio are probably available on Facebook, Instagram, TikTok, and the like. But it’s worse for people in the public eye. People on local government councils or boards may have their voices online due to being involved in public meetings.
Another way scammers can get a hold of your voice is by calling you directly, engaging you in a conversation, and then piecing together a different conversation based on what they’ve recorded from a seemingly weird or random phone call on your end. The best thing to do if you get a suspicious phone call is hang up!
Protect your bank account
If you want to be certain that your bank account, or your municipal bank account, is safe from this type of scam, go to the bank and tell them not to accept instruction over the phone. While this could be a bit of an inconvenience if you need to make a change to an account, it could end up saving you time and money in the long run.
How to tell if a phone call is fake
It can be difficult to tell if a phone call is real or fake, especially when scammers try to make it a high-stress situation. Try to keep your wits about you and remember these five warning signs of an AI voice scam:
- You only briefly “hear” your loved one’s voice. – The longer a scammer uses a cloned voice, the more of a chance there is you’ll hear that slightly robotic sound and catch on. Therefore, many scammers will play only a brief clip when trying to scam idividuals.
- They can’t answer simple questions. – The scammers don’t know you, your memories, your silly quirks. Try to pry these things out of the person on the other end. If it’s a scam, they will not be able to produce.
- You’re called from an unknown number. – It’s very easy to spoof a number. This may or may not be the case. But it can be a clue to a scam call.
- Someone else quickly takes over the call. – Scammers want to get something from you. Pretending to be a lawyer, law enforcement, kidnapper, or some one else is usually the quickest way to the goal.
- You’re told to pay a ransom via cryptocurrency or gift cards. – This is a recurring theme with scammers. It’s much more difficult to get your money back when you pay with cryptocurrency or gift cards. Scammers know this and take advantage.
