SecurityBrief New Zealand - Technology news for CISOs & cybersecurity decision-makers
Story image
AI deepfake scams surge with celebrity voice cloning on social media
Wed, 28th Feb 2024

Advanced AI technology has increased the frequency of celebrity deepfake scams, particularly audio scams on popular social media platforms. Bitdefender's research signals the rapid growth of audio deepfake scams published on platforms such as Facebook. These scams use voice cloning technology to replicate the tone and speech patterns of notable celebrities, including Oprah, Jennifer Aniston, Elon Musk, and YouTube star Mr. Beast.

Bitdefender's Security Analyst, Alina Bizga, highlights the typical scams that involve celebrities endorsing fake giveaways, gambling, or investment opportunities. A particularly insidious scam includes a video promoting a time-limited giveaway of popular products, such as iPhones and MacBooks. Victims are told they will receive one of these products for free if they just pay for shipping. They are persuaded to enter personal details and credit card information to secure their so-called prize.

Voice cloning, a remarkable achievement of AI, creates synthetic copies of an individual's voice. It utilises deep learning techniques to generate a highly realistic imitation of human speech. Alina Bizga explains the process: voice data is collected, analysed for unique characteristics, used to train a machine learning model, and ultimately converted to synthetic speech. By continuously supplementing the AI with more voice data and fine-tuning parameters, the synthetic cloned voice becomes more convincingly human.

While the process possesses lawful uses, including education and healthcare applications, voice cloning technology is invariably misused for fraudulent activities. Bitdefender's research reveals that cyber criminals applying AI voice cloning are causing financial and physical harm to unsuspecting victims. Fraudsters masquerade as victims' relatives, extort money through virtual kidnapping schemes or use social engineering schemes to trick users into transferring money or accessing malicious links.

Scammers exploit the vast userbase of social media platforms for their voice cloning scams. With millions of potential victims, personal data is easily obtainable, and the global reach is extensive.

Bitdefender Labs discovered fraudulent content using AI voice generators to clone the voices of several well-known celebrities. The false ads lured potential victims with promises of exclusive iPhone, MacBook and other high-end product giveaways for minimal shipping costs. These scams were further propagated through fraudulent Facebook ads redirecting them to scam websites promoting giveaways or investments. Often, upon clicking on the advertisements, users were redirected to lookalike websites of popular news outlets, adding another layer of credibility to the fraud.

The deepfake videos used in the scams were often poorly executed, with clear artefacts and inconsistencies. Despite this, they reached at least 1 million US and European users, primarily individuals aged between 18 and 65. The scam, however, is not delimited to these geographical zones; it is likely a global issue tailored to individual consumer trends.

Bitdefender encourages users to stay vigilant against these scam tactics. This includes scrutinising the quality and consistency of voices in media, being wary of unusual requests and "too-good-to-be-true" deals, verifying suspicious information with the supposed source, using comprehensive security solutions such as Bitdefender, and staying informed about the latest scam tactics.