Alert! AI-based voice cloning is the latest cybercrime

Litmus test for humans as GenAI starts taking away entry-level jobs

Srinagar: Cyber Police has asked the people to remain alert as the scammers are now using Artificial Intelligence (AI)-based applications to mint money from “gullible people”.

The increasing use and acceptance of AI tools have made it simpler to manipulate images, videos, and voices of acquaintances and relatives. Recently, news surfaced that cyber attackers are exploiting AI-based voice technology to deceive individuals.

   

Fraudsters are taking advantage of AI to mimic the voices of distressed family members, and a considerable number of people are becoming victims of such scams.

Recently a report by McAfee suggested that a majority (69 percent) of Indians are unable to distinguish between a genuine human voice and an AI-generated voice. Since each person’s voice is distinctive, it can be considered as a biometric fingerprint that establishes credibility.

However, the prevalent practice of 86 percent of Indian adults sharing their voice data online or through recorded notes at least once a week (on social media and through voice notes) has made voice cloning a potent weapon for cybercriminals.

The advisory by J&K Police said that AI-based voice cloning is the new cybercrime in news. “Apple Ios7 and various apps and sites like Heyge, Murf, and Resemble Al are being used for voice cloning,” the advisory reads. It said that the criminals were using computer-generated voice sounds to act as friends and relatives, and in the process, money could be lost.

Sharing the modus operandi, Police said that the AI-ML (Artificial Intelligence and Machine Learning) model learns the patterns and characteristics from the person’s voice recordings (speech patterns, accents, voice inflection, and even breathing).

Police said that the learning includes pronunciation of words, tone of the voice, and emotions.

Now criminals use this AI-ML model to convert text to speech as it generates a waveform to represent the sound of any person.

The advisory said that the criminal then uses the computer-generated call to call his or her relatives or friends and based upon the familiarity of the voice, the victim ends up paying the money.

“The receiver should listen carefully to unnatural pause or robotic speech styles, look for errors in pronunciation or speaking tone,” it said. “The voice can be compared with the original person and notice if the voice lacks emotional expression or difference.”

Leave a Reply

Your email address will not be published. Required fields are marked *

four − 2 =