Protecting Your Business from Deepfake Deception

Deepfakes — digital forgeries produced by artificial intelligence (AI) — have blurred the line between reality and illusion. On the upside, AI-generated deepfakes have revolutionized special effects in motion pictures and made certain education and health care industry processes more effective. Yet there are also plenty of risks associated with deepfakes.

Current Threats

Deepfakes purporting to represent public officials can disseminate disinformation and generate fake news stories. And if fraud perpetrators use deepfake images of a company's owner or senior executives, they can more easily perpetrate phishing schemes and steal sensitive data.

The threat extends beyond visible manipulation to audio. Deepfakes can mimic a specific individual's voice to commit theft. For example, a so-called "business partner" might leave a voicemail instructing someone in your accounting department to wire funds to an overseas account.

Detection Challenges

AI-based detection technology solutions can help reveal deepfakes by identifying unusual facial movements, unnatural body postures and lighting inconsistencies. Yet this technology is still in its infancy and far from perfect.

Alternative solutions, such as watermarking, show promise. However, watermarking technology is relatively easy to bypass and has yet to gain widespread acceptance. A small but growing body of law regulates the use of deepfakes. But the laws do little to prevent their creation. They generally punish creators when (and if) they're caught using deepfakes to commit illegal acts.

Key Warning Signs

Recognizing the red flags of deepfake content is vital. You and your employees should be wary of video or audio exhibiting:

• ‍Unnatural eye movements. Deepfake creators find it particularly challenging to replicate natural blinking patterns, eye movements and eye gazes. Inconsistencies in eye-related movements could be suspicious.‍
• Unrealistic faces. Mismatched skin tones, questionable lighting and blurred edges are potential signs of a deepfake. So, too, are stiff or exaggerated facial expressions.‍
• Lip-sync and audio issues. Lip movement lagging its soundtrack is a common problem that's difficult for deepfake creators to overcome. A deepfake may not be able to capture an individual's tone or emotion and its soundtrack may contain abrupt or unnatural pauses.‍
• Corrupted backgrounds. Warped objects near the edges of a person's face or inconsistent backgrounds that bleed into the foreground are possible signs of a deepfake.‍
• Unbelievable content. Deepfake videos can be entertaining, but they're also frequently sensational and out of character for the individuals supposedly being depicted.‍
• Questionable sourcing. Many deepfakes are from non-credible sources and circulated via untrustworthy platforms. Any content that goes viral, meaning people share it with their social networks, should be treated with caution.

Corroborate Files First

Until technology makes it easier to uncover deepfakes, exercising a healthy skepticism is the best way to avoid being conned. Before you treat a video or audio file as legitimate, corroborate it with multiple sources. And if employees receive an unusual request via voicemail or video from a supposed manager, they should verify it by phone or by talking to the individual in person.

Contact us with questions and for help training your workers to fight malicious deepfakes and other fraud schemes.