AI-backed Deepfake Impersonations Are Getting Harder to Detect, FBI Warns

AI-backed Deepfake Impersonations Are Getting Harder to Detect, FBI Warns

Naveen Athrappully

Increasingly hard-to-detect deepfake content created with artificial intelligence is being exploited by criminals to impersonate trusted individuals, the FBI and the American Bankers Association (ABA) said in a report published on Sept. 3.

In its “Deepfake Media Scams” infographic, the FBI said that scams targeting Americans are surging. Since 2020, the agency has received more than 4.2 million reports of fraud, amounting to $50.5 billion in losses. “Imposter scams in particular are on the rise. ... Criminals are using deepfakes, or media that is generated or manipulated by AI, to gain your trust and scam you out of your hard-earned money.”

Deepfake content can include altered images, audio, or video. Scammers may pose as family, friends, or public figures, including celebrities, law enforcement, and government officials, the FBI warned.

“Deepfakes are becoming increasingly sophisticated and harder to detect,” said Sam Kunjukunju, vice president of consumer education for the ABA Foundation.

According to the infographic, certain inconsistencies in the AI-generated material can help detect deepfakes.

When it comes to images or videos, people should watch out for blurred or distorted faces; unnatural shadows or lighting; whether audio and video are out of sync; whether the teeth and hair look real; and whether the person blinks too little or too much. In the case of audio, people should listen closely to determine if the tone of voice is too flat or unnatural.

More...