Deep fakes are becoming a major concern worldwide, with everyone from politicians to executives worried about this new, AI-generated threat. As experts in cybersecurity in Miami, it’s crucial that our clients know about deep fakes and how to spot them.
What are Deep Fakes?
Deep fakes use artificial intelligence to engage in what is known as “deep learning.” Through the use of AI, it’s possible to create videos by analyzing thousands of videos and pictures to create extremely realistic fakes.
The complex algorithms will look for similarities in faces and shared features using a decoder. Multiple decoders may be used at one time to swap faces and voices. After the technology has “learned” enough, it will recreate the person’s face and movements with impeccable realism.
Using AI, it’s possible to digitally alter videos to:
- Spread false information
- Act maliciously
A video message from the CEO of a company can be sent to the security team via email to ask for access to a company’s database. In this scenario, the threat of a deep fake can have significant security implications.
Politicians have already been the target of deep fakes, which were seen by millions of people before verifying that they were fake.
With deep fakes, it’s possible to create synthetic media, causing a major trust issue in both the media and business worlds.
How You Can Spot Deep Fakes
Initially, spotting deep fakes was easier than it is now. Some videos would have slight stuttering in the images or things seemed off with the fakes. For example, since images may be used to create a deep fake, the videos they created would not have a person blinking, but the AI behind these fakes has advanced drastically in recent years.
By the time research is released on how to spot deep fakes properly, there are people dissecting the information and updating deep fakes to address researchers' findings.
A few of the ways to detect deep fakes of poor quality include:
- Strange lighting that doesn’t match up with the rest of the video
- Odd reflections in the person’s iris
- Lips not syncing up with the words being said in the video
- Odd flickering near the person’s face
In fact, many of the world’s largest tech companies are working on solutions to spot deep fakes. Contests were held to detect fakes, and companies such as Amazon, Facebook and Microsoft were all involved in the process.
Deep fakes have advanced so much in the past four years that most people cannot spot a “good” fake on their own. Utilizing the latest detection systems is a good start, but the accuracy depends on the images available.
Detection systems rely on analyzing thousands of images to spot fakes.
If a politician or CEO is rarely photographed, the accuracy of modern detection systems will continue to decline. Celebrities are the easiest to discern a fake from a real image because there are often thousands of images to help detection systems learn how to spot fakes.
However, if you’re live calling someone and a deep fake is used, one method of detection is asking the person to show their profile to the camera. Otherwise, without the most advanced detection systems, it can be painstakingly difficult to tell a fake from a real video or image.