Rise of Celebrity Deepfakes – How Technology is Transforming

Social media platforms have become an integral part of our daily and fraudsters use these platforms to perform their fraudulent endeavors. Celebrity deepfakes are trending on social media platforms like Instagram, Facebook, Twitter, TikTok, and YouTube, aimed to damage celebrity’s influential image. A study reveals that out of 4,000 deep fakes, nearly 255 are British celebrities including TV actors, musicians, and YouTubers. In 2023, almost 43,733 cases of deep fakes were reportedly posted on top 40 fleshy websites, surpassing the record of all previous years.

What are Deep Fakes?

The term Deep fakes is a fusion of two words ‘deep learning’ and ‘fake media’ and Generative Adversarial Networks (GANs) are integrated into generating deep fakes. GAN models leverage convolutional neural networks (CNNs) to recognize patterns in the data and generate new identities resembling the genuine ID. Deep fakes may be fake images, videos, or even voices of individuals, obscuring thresholds between real and fake identities. Malicious actors have become sophisticated with advancements in technology and use nefarious ways to upgrade their fraudulent tactics.

Celebrity deepfake are experiencing a sharp rise, in 2016, just a single exclusive deep fake was reported, however, 2023 witnessed the highest cases of fleshy deepfakes compared to previous years. The question is how deepfakes are generated and why identity verification solutions are not able to detect them. The answer is simple, malicious actors have sophisticated their morphing techniques, used for the deep fake generation, that even dodge active liveness detection.

Fraudsters acquire biometric data of individuals from different sources like digital footprints, social media platforms, and identity documents, and leverage the procured data to generate a completely new ID identical to the victim’s ID.

Emergence of Deep Fake Technology

The history of deep fakes can be traced back to the 1990s when research was conducted to analyze how artificial intelligence could be implemented in the manipulation and evaluation of images. The technology was not utilized for the projected goal until the middle of 2010 and it gained popularity in 2014 with the evolution of neural networks and the introduction of GANs by Ian Goodfellow. GANs lay down the base for the generation of deep fakes, which are utilized to mimic real persons and get unauthorized access to multiple services.

In days gone by, it was challenging and arduous to generate deep fakes, but with technological advancements, the process has become effortless leveraging AI algorithms. Back in November 2017, a Reddit user developed a video-editing software enabling potential users to train their computers. The software was able to seamlessly swap adult actor IDs with celebs IDs, consequently used for diverse purposes. Thereafter, many deep fake websites and applications have been developed to create celebrity deepfakes.

Instances of Celebrity Deepfakes

Celebrities are the most obvious targets for deep fakes, however, the seamless implementation of deep fake technology reveals the whole society is at risk of AI misuse. Fraudsters generate deep fakes to exploit the reputation of influential people like politicians, corporate leaders, state heads, and journalists. The manipulated data is utilized to spread misinformation, deceive the public by portraying wrong information, and defame individuals, leaving far-reaching consequences on personal as well as communal levels.

A few examples of celebrity deepfakes include

Deep Fake Images of Taylor Swift

AI deep fake is so advanced that it’s hard to differentiate between real and fake identities. Back in January, Taylor Swift, an American singer, and Spotify’s most-played artist, fell victim to deep fakes. The images of Taylor Swift circulated on X, formerly known as Twitter, for nearly 19 hours and crossed more than 45 million views, spurring X to block all search items to remove search terms related to this incident. The incident caught the White House’s attention, pushing Congress toward establishing stringent regulations to penalize vicious and fake images.

Deep Fake Video of Tom Hanks

A manipulated video of Tom Hanks, an American Oscar winner actor, was roaming across social media platforms, promoting his dental plan. The deep fake video circulated on social media for many days, on 21st October, Tom Hanks took to Instagram and alerted his 10M followers that the advertisement was made without his consent.

Voice Cloning of Scarlett Johansson

Scarlett Johanson, the world’s highest-paid actress in 2018 & 2019, fell victim to a recent deep fake advertisement where her name and voice were manipulated to promote an unauthorized firm. AI algorithms were integrated to generate manipulated advertisements for a firm called ‘Lisa AI’ using Scarlet Johansson’s deep fake. Eventually, it was confirmed that she was not involved in app promotion and her lawyers vowed to take legal action against this.

Final Thoughts

Not all celebrity deepfakes are generated with harmful intentions, some forms are used for entertainment & media production. In some instances, AI deep fakes are utilized to de-age elderly or senior actors and used in filmmaking. However, we can’t unsee the negative consequences of deep fakes tormenting the reputation of celebrities and high-influential individuals. The identity verification tools must use advanced AI algorithms and train them on extensive datasets to identify fake IDs. Advanced IDV solutions must leverage facial recognition technology and liveness detection to accurately verify the ID by minimizing False acceptance & False rejection rates. Additionally, efforts must be made to enhance the speed and accuracy of such tools for effective performance. Detecting deep fakes could be challenging, however, it’s not something impossible. Analyzing facial cues, identifying anomalies, using search tools, and staying cautious are some major steps to detect ever-evolving deepfakes.