Many of us have seen celebrities, leaders, and sportspersons flagging disturbing videos where they have been seen promoting brands, cursing, and jogging. They are deepfakes or to put it simply AI-generated fake videos. It raises an alarm, but only when used negatively. In this digital age, such a technology can help answer many queries.
“Deepfake is a mechanism to generate fake audio, fake image, and fake video through the use of artificial intelligence,” cyber security expert Rafay Baloch says in an interview with Aaj News.
It is far better than Photoshop or video editing, making it difficult for even a person from a background in graphic design to differentiate between a fake and an original picture.
What we have seen in the recent past is that nothing is sacrosanct. With the help of the latest software, only one to two pictures are needed to create a deepfake as compared to the past when thousands of pictures were needed to have fake visual content.
“No one has the answer for how to be safe from the deepfake. You cannot prevent somebody from creating your deepfake,” Baloch stated. But he shared that Facebook, Google, and top research institutions are working on its detection.
The world is yet to create a detection mechanism or software to 100 per cent detect deepfake, the cyber security expert added.
While speaking about the positive impacts of deepfake technology, Baloch shared an example where an Indian politician used deepfake to translate his speech into seven languages such as videos in their election campaign.
He mentioned Hollywood actor Henry Cavill who was working on two projects, Batman v Superman: Dawn of Justice and Mission: Impossible - Fallout at a single time. Both the films were released in 2016.
Baloch shared that the British actor was required to have a moustache in the Tom Cruise-starrer movie while producers wanted a clean shave look in the Warner Bros picture. So, to address this issue the filmmakers used computer-generated imagery (CGI) to show Cavill without a moustache, triggering criticism from fans. Later, a person claimed that he created the same character with the help of deepfake software in five minutes. He spent $1,000 on the project compared to the filmmakers who spent $25 million to present the iconic superhero.
Similarly, many would recall how artificial intelligence and CGI were used to make Cody and Caleb as Paul Walker in the Fast and Furious 7. The two brothers of the late US actor acted in the movie to complete the project.
He went on to add that the artificial intelligence which enables such a technology has a big scope, adding that machine learning and cyber security were among emerging technologies.
Baloch, who is known for his discovery of vulnerabilities on the Android operating system, said that top world economies categorises people with such skills differently and offer different visa facilities.
Also, read this
Fact-Check: Don’t get fooled by these viral images generated by AI Tools
Many actors have already signed contracts with companies, allowing the production houses to make and use their deepfake after their death. The royalties of which will go to our family.
But cyber security experts believed that jobs would not increase but decrease after the advent of AI technology as it reduces the number of people needed to complete a task.
According to Baloch, an artist can file a copyright claim if their face or voice is used without their permission.
“If someone’s photos go viral, he can contact the Federal Investigation Agency and the Pakistan Telecommunication Authority and take action against him. Individually, affected people can also approach institutions via social media applications,” he said.
When asked how deepfakes can be detected with the help of AI, he explained that deepfake are being created when two AI algorithms fight each other. He added that one creates it and another detects it.
But there are some tips by which such videos can be detected.
“First of all, the blinking of the eye in deepfakes is not at the same rate as human eyes. You will see irregularities in fake videos. So far softwares have been notable to solve this issue,” he said, “there are anomalies in lip-syncing. You can detect through that,” he said.
“The imagery of deepfake is shot from the front, not from the side, as it is difficult for deepfake to create pictures from sides. This is also an indicator which you can see.”
