AI is impacting nearly every sector including fraud. Generative AI technology has also spawned increasingly sophisticated methods of deception, with deep fake technology emerging as a significant threat in South Africa.
Criminals are creating more advanced versions of impersonation fraud using AI-generated video and voice. These are so convincing that businesses have been duped into transferring millions to scammers, like an employee at the Hong Kong office of a multinational company who transferred $26 million after unwittingly attending a video call with deep fakes of co-workers, including the company’s chief financial officer.
A Luno staff member was recently targeted using a deep fake audio message on WhatsApp. Purportedly from a Luno executive, the instruction in the voice note raised flags and alert staff were able to avert a problem.
Deep fakes pose a material risk in the crypto sector too, according to Johan Hetzel, global head of compliance and anti-financial crime at Luno, a licensed financial services provider and South Africa’s largest crypto investment app. “The blockchain (the technology that powers cryptocurrencies) allows users to track most transactions. Since the blockchain is hosted publicly, crypto service providers can trace transactions beyond their customer base.
This differs from traditional finance, where fraud syndicates distribute funds between different financial institutions, making it difficult to track transactions. But it’s human nature to believe what we see and hear for ourselves and deep fake videos and audios are not easy to identify,” he says.
Hetzel unpacks how to spot deep fakes and how Luno is leading the way in protecting customers.
Countering deep fakes
Elon Musk is a favourite subject for fraudsters. Not only is he one of the richest businessmen in the world, Musk has discussed crypto on his social media accounts. During the recent launch of SpaceX’s Starship rocket, YouTube was flooded with deep fake videos of Musk promising double returns to investors for making deposits into a fake scheme.
Luno has a dedicated team focused on keeping customers safe. It employs various tools to scan domains and social media to find instances where Luno’s brand is used in fraudulent websites or social media pages. The approach focuses on removing these platforms fast.
Over the last year, Luno has proactively detected over 2,000 fake social media sites with a takedown success rate of 98% and almost 500 fake websites, with a takedown success rate of 87.6%.
Red flags
One of the first warning signs of a deep fake is the communication channel used. Consider how you were contacted – if you received an email, look carefully at the email address to confirm that it is legitimate. If you were contacted through social media, how recently was the account created, how many friends or contacts does the person have? If you know the person who reached out to you, contact them to confirm whether they did in fact request something from you (a payment, for instance) and whether this request is genuine.
Unnatural facial expressions or facial morphing can be a sign of a deep fake. You can also check for faces that are distorted, or skin that looks too smooth or wrinkled. Other visual cues in deep fake videos are unnatural eye movements, such as no blinking, shadows in unexpected places around the eyes and eyebrows and hair that doesn’t look real.
AI-generated audio is harder to identify because you rely solely on your hearing, but there are clues. AI voices generally don’t display much emotion and tonal variation. Does the speaker take a breath before speaking? Also pay attention to whether there are any background noises.
Blockchain technology
Blockchain technology can establish authenticity for content and information – due to its transparent, decentralised nature and its focus on network security and encryption. Blockchain has already demonstrated the ability to authenticate content with digital art such as non-fungible tokens (NFTs). Blockchain tech allows anyone to verify the creator and owner of a piece of art, making it possible to distinguish between the original and potential replicas. This transparency and authentication potential extends to videos, images, and text.
General tips to avoid being scammed
Says Hetzel, “Fraud is a concern for crypto service providers, as well as across the financial services industry in South Africa. To stay safe, remember that if it’s too good to be true, it probably is. Be extra careful (even paranoid) and verify the person or organisation you are dealing with by doing extensive online research and asking them specific questions about their business. Fraudsters often rush you into making hasty decisions. They are very persuasive! Take five minutes to think things through. Don’t be afraid to terminate a conversation and walk away.”
