Cutting Through The Clutter: Real Intelligence on Deepfakes

Written by Don Lariviere

‍

Every day more of us get our morning wake up from Siri or Alexa, the news and weather from our Google Assistant, and begin to turn our lights on just from wishing it out loud. We can now have complete conversations over our morning coffee and in any language we’d like with Chat-GPT4o. 

That doesn’t even scratch the surface of all of the ways that artificial intelligence (AI) is entering more aspects of our lives. But for all of these and an endless list of great things that AI brings today and tomorrow, we’re also seeing the concerning invasion of deepfakes. 

Deepfakes have emerged as a significant and sophisticated phenomenon. These very realistic but fabricated videos and other types of messages, including audio, use AI to create original and highly life-like content. 

So What Are Deepfakes?

Deepfakes are videos or images that have artificially transformed a person’s words, speech, or likeness. The term deepfake combines "deep learning" and "fake," reflecting the technology and its deceptive nature. 

Deep learning, a subfield of machine learning, involves training neural networks on vast amounts of data to learn the patterns and intricacies of data to achieve superhuman capabilities in processing, recognition, and generation. These algorithms can generate highly convincing synthetic images, audio, and videos.

The technology behind deepfakes involves two key components: the generator and the discriminator. The generator creates fake images, while the discriminator attempts to detect if an image is real or fake. Through this adversarial process, known as a Generative Adversarial Network (GAN), the system continually improves, producing increasingly realistic deepfakes.

Why Are Deepfakes So Prevalent?

The rapid development of AI and machine learning has made these types of image and video manipulation readily available. Tools and software for creating deepfakes can be free or very inexpensive, and there’s clearly so much data to fill them! The widely used Laion 5B is such a dataset.

The sheer vastness of data available online, including images and videos, provides lots of training material for AI algorithms. Social media platforms and other digital repositories are rich sources of personal data, making it easier to create convincing deepfakes.

Initially, deepfakes gained popularity in the entertainment industry for creating special effects and realistic simulations. Their novelty also attracted amateur creators, leading to a surge in deepfake content across the internet.

In today’s political climate, deepfakes are also being weaponized for disinformation campaigns. The ability to create realistic but false content has significant implications for public trust and safety. 

Of course there are plenty of widely acceptable and legitimate uses for deepfakes as well. Many of the companies and services we interact with have used synthetic, real-life avatars for our routine and complex customer service inquiries, for training and developing their new employees, even adaptable education experiences to create accessibility for students. AI-enabled devices remind us what we’re probably shopping for when we step into our local grocery store at 5 p.m. on a Tuesday, or pre-order our favorites when we’re five minutes away from the coffee shop. 

But…They Can Be Illegal 

For all those life-changing perks, deepfakes can also pose risks that are or could be illegal. While the United States doesn't have a federal law, some states do, as do many countries in the European Union and elsewhere around the globe. 

Deepfakes can be used to impersonate people, leading to fraudulent activities such as unauthorized financial transactions, identity theft, and harassment. 

Again, they can be employed in political and social contexts to spread false information, manipulate public opinion, and destabilize institutions and relationships. They have been used to create fake news videos that can sway voting decisions or incite violence. Deepfakes have been used in phishing attacks, where voice or video messages from trusted sources are mimicked to gain access to sensitive information or systems.

How To Spot Deepfakes

Deepfakes are getting more difficult to detect as the sophistication level of criminals spikes. Identifying deepfakes requires a keen eye and awareness of just a few of the telltale signs.

1. Inconsistent facial features: Look for unnatural facial movements such as unusual blinking patterns, inconsistent eye alignment, or mismatched lighting on the face compared to the rest of the scene.

2. Audio-visual discrepancies: Listen for audio mismatches, such as bad lip-syncing or weird speech patterns. Deepfakes often struggle with replicating natural speech nuances, patterns, accents, natural pauses, and pronunciation. 

3. Artifacts and blurring: Deepfakes may contain visual artifacts, blurring, or pixelation, especially around the edges of the face or in complex scenes.

4. Unnatural body movements: Deepfakes might exhibit awkward or unnatural body movements that don’t match up with the realistic portrayal of human behavior. Also watch a person’s hair — inconsistencies in hair motion (moving when it shouldn’t be), or very blurry hair is often a tip-off.

What Should We Do If We Think We’re Being Duped?

1. Verify the content: Before reporting, try to verify the content through multiple sources to ensure it is indeed a deepfake.

2. Contact platforms: Report the deepfake to the platform where it was found. Social media sites and video platforms often have specific channels for reporting deceptive content.

3. Notify authorities: In cases involving serious threats or illegal activities, notify local law enforcement or relevant authorities who can take appropriate action.

4. If detected at work, notify the right people: Let your IT, Information Security, Data Privacy, and/or Legal teams about the suspected deepfake. Businesses of all sizes are often the target of cybercriminals, and Proxet can partner with you on combating them. 

5. Seek legal advice: If you are a victim of a deepfake, consider seeking legal advice to understand your rights and potential remedies.

Conclusion

Deepfakes represent a significant challenge in the digital age, blending the benefits of technological innovation with the harmful potential for misuse. Understanding what deepfakes are, recognizing their prevalence, and being aware of their illegal uses are critical steps in combating this threat. 

Through a combination of detection tools, legal measures, and public education, we can all better safeguard ourselves against the risks posed by deepfakes!