Businesses and societies benefit a lot from technological progress. But progress brings with it new risks that are hard to deal with. Artificial Intelligence (AI) is one of the most important new technologies. It is being used in more places than ever before.
AI has a huge amount of business potential. It can do everything from automating clerical tasks to finding hidden business drivers. But bad AI use can hurt businesses and cause them to lose a lot of credibilities.
The FBI recently talked about a growing trend caused by the rise of remote work, in which bad people used deepfakes to pose as job applicants in American companies. These people stole the identities of U.S. citizens in order to get into company systems. It has a huge effect on corporate espionage and security.
How can companies stop the use of deepfakes, which is on the rise even as the technology behind them gets better? Here are some ways to make security risks less dangerous.
Table of Contents
Most of the time, the best way to deal with advanced technology is to go back to the basics. Deepfakes are made by stealing a person’s identifying information, like pictures and ID numbers, and then using an AI engine to make a digital copy of that person. Malicious people often use existing video, audio, and graphics to imitate their victim’s speech and actions.
A recent case showed just how far bad people will go to use this technology. Several European political leaders thought they were talking to Vitali Klitschko, the mayor of Kyiv. They were later told that they were talking to a deepfake.
After calling the Ukrainian embassy and finding out that Klitschko was busy elsewhere, the office of the mayor of Berlin found out about the plan. Companies would do well to think about what they can learn from this. Using deepfakes can be found out through identity checks and other seemingly simple checks.
When interviewing candidates for remote jobs, companies run the risk of deepfake encounters. Rolling back the rules on remote work is not a good idea for companies that want to hire the best people. But the chances of hiring a deepfake actor will go down if you ask candidates to show some kind of official ID, record video interviews, and make new hires come to the office at least once right away after being hired.
Even though these methods won’t stop deepfake risks, they will make it less likely that a bad actor will get access to company secrets. Like how two-factor authentication stops bad people from getting into systems, these analog methods can stop people from using deepfakes.
Verifying an applicant’s references, including their picture and name, is another analog method. For example, send the authority a picture of the applicant and ask if they know that person. Check the reference’s credentials by talking to them in a business or official setting.
Fight fire with fire
Deep learning (DL) algorithms are used by Deepfake technology to copy a person’s actions and habits. It can lead to scary things. With just a few data points, AI can make moving pictures and videos that look like they were made by us.
Deepfakes can be stopped with analog methods, but it takes time. Using technology against itself is one way to find deepfakes quickly. If DL algorithms can be used to make deepfakes, why can’t they also be used to spot them?
In 2020, Maneesh Agrawala of Stanford University came up with a way for filmmakers to add words to the sentences of people they were filming. From the outside, nothing looked wrong. Filmmakers were happy because they wouldn’t have to reshoot scenes because the sound or dialogue wasn’t right. But this technology caused a lot of bad things to happen.
Agrawala and his team knew about this problem, so they made another AI-based tool that could find differences between how the lips moved and how the words were said. Deepfakes that add words to videos in a person’s voice can’t change how their lips move or how they look.
Face impositions and other common deepfake tricks can also be found with Agrawala’s solution. As with all AI applications, a lot depends on what information is given to the algorithm. But even this variable shows that there is a link between deepfake technology and the way to stop it.
Deepfakes use made-up data and datasets that are extrapolated from real-world events to cover a wide range of situations. For example, artificial data algorithms can use information from an incident on a military battlefield to create more incidents. These algorithms can change things like the state of the ground, the readiness of the participants, the state of the weapons, etc., and feed that information into simulations.
Companies can fight deepfake use cases with this kind of fake data. AI can predict and find edge use cases by extrapolating data from current uses. This helps us learn more about how deepfakes are changing over time.
Accelerate digital transformation and education
Agrawala says that there is no long-term solution to deepfakes, even though the technology to stop them is very advanced. At first glance, this seems like a sad message. But companies can stop deepfakes by speeding up their digital strategies and teaching their employees the best ways to do things.
For example, deepfake awareness will help employees look at the information and figure out what it means. Any information that is going around that seems crazy or out of place can be called out right away. Thanks to deepfake threats, companies can make procedures for verifying employees’ identities when they work from home and be sure that their workers will follow them.
Again, these ways can’t deal with deepfake dangers on their own. But using all of the above methods, companies can build a strong framework that makes deepfake threats less likely.
Advanced tech calls for innovative solutions
The best way to stop deepfake threats is to make technology better. In a strange way, the answer to deepfakes can be found in the technology that makes them work. There will definitely be new ways to fight this threat in the future. In the meantime, businesses need to be aware of the risks deepfakes pose and work to reduce them.