Artificial Intelligence may be the solution for many problems, but when it’s used with bad intent, it can pose serious threats. Deepfakes, for example, now represent a threat against truth that is evolving and strengthening at an alarming rate. Deepfake technology is a step ahead of current preventative and identification measures, so it’s imperative we learn to identify and combat it in real time. Here’s what you need to know.
What They Are
Deepfakes are falsified images, video and audio that depict real people, places or events in damaging ways. They go far beyond Photoshop, and almost anyone can create them. They can digitally place your CEO, general counsel or board member in compromising situations, depicting them uttering sexist, racist or otherwise outrageous statements. A notable example from 2018 shows a fake Barack Obama insulting the current president. Since then, deepfakes are harder to discern from reality and have been damaging reputations of politicians and corporate executives.
You begin with Generative Adversarial Networks. GANs are specialized machine learning systems that take real data (i.e. thousands of photos of Barack Obama) and create convincing replicas to produce fake images and video. The same algorithms are applied to audio imitation.
How to Identify Them
Because the algorithms used to create deepfakes are based on real data, even machine learning systems struggle to discern between the real and the fake. For some of the less sophisticated examples, though, systems can spot tell-tale signs, like unnatural blinking, head positioning and speech cadence. However, deepfake technology is only getting more sophisticated, and deepfake developers have access to the same identification methods as those fighting them. Facebook, Microsoft and a group of academics are building a database to fight deepfakes.
Common Attack Tactics
So, what does all this mean for communicators? A start is raising awareness around the most common ways you’ll encounter deepfakes firsthand:
- Phishing attacks use impersonation tactics, like familiar-but-off-by-one-letter email addresses, to fool targets into divulging sensitive data or transferring funds. Deepfakers take phishing a step furthering with whaling by impersonating your CEO, for example.
- Hackers will use digitally created images, video and audio of industry leaders, politicians, and government officials to spread false information, such as fake stock market prices.
- Another potential threat from deepfakes is the increased likelihood and effectiveness of blackmail and extortion attempts.
How to Prepare
To insulate your business and companies you advise from the fallout of deepfakes, the most important step is to include a deepfake scenario as part of your crisis communication plan.
You’ll want to:
- Create a process for verifying deepfakes used against you.
- Determine how to communicate with social media platforms so the deepfake can be removed.
- Plan to communicate with your employees, customers and other stakeholders to minimize reputation damage.
In addition, use these countermeasures:
- Spread awareness and education on deepfakes among colleagues and those companies you represent and adopt security training and protocols.
- Minimize systems used for communications; use them consistently so that outliers can be easily spotted.
- Use listening software to monitor your brand’s online presence, enabling you to know about a potential deepfake in real-time.
- Deepfake technology is outpacing detection technology. Still, detection technology is superior to relying on human detection. Having an AI report bolsters the ability to defend your deepfake claim.
Ellen Huber is an account supervisor at kglobal, LLC