Beware the deepfake con artists


From the fun Tom Cruise impersonation, to the more concerning fake Nancy Pelosi video, “deepfakes” have gained a lot of attention. But what are they and why could they be harmful to your business?


How are deepfakes made?


“Deepfakes”- a mashup of “deep learning” and “fake” – leverage techniques from Machine Learning and Artificial Intelligence to create fake video and audio. 


Convincing fakes are still difficult to make as they require sophisticated technology. Trying to replicate a well-known person is particularly challenging as there are subtle differences in intonation or delivery that can reveal a fake. The reason the Tom Cruise deepfake worked so well was because a skilled impersonator was already able to closely mimic Cruise’s mannerisms. The AI simply ‘blurred’ the audio and visual data to align it even more closely to the real Tom Cruise, using algorithms known as generative adversarial networks (GANs).


What are the dangers?


As AI improves, deepfakes will be harder to spot and their use in criminal activity is likely to increase. We’re already seeing them used in pornographic videos as a method of blackmail, and in elaborate heists to con businesses out of millions of dollars. They also have the potential to be able to sway elections and cause political instability. 


How can businesses protect themselves from deepfakes?


For now, the most common way to protect your business from deepfake con artists is to examine the circumstances around interactions – the time, place and method of the communication – rather than trying to identify the deepfake itself. This is similar to how mobile carriers flag potential spam phone calls or phishing texts. Microsoft, however, has developed a video authenticator to detect deepfakes as a way of combating disinformation. 


Of course, none of this technology is 100% effective against deepfakes, so businesses today need to remain vigilant against any suspicious activity.