Deepfakes: The Rise of Risky Synthetic Media and Its Implications

New Thread

Ctrl

Home

Discover

Try Pro

Upgrade for image upload, smarter AI, and more Pro Search.

Introduction
Types of Deepfakes: Identifying the Dangers
How Deepfake Works: The Technical Process of Switching Faces
How to Recognize a Deepfake
How to Keep Yourself Safe From Deepfakes
Closing Thoughts on Deepfakes

Deepfakes: The Rise of Risky Synthetic Media and Its Implications

Curated by

cdteliot

5 min read

137

Deepfakes, a form of AI-generated synthetic media, have proliferated rapidly in recent years, raising concerns about their potential to spread misinformation and erode trust in digital content. According to a study by Harvard's Misinformation Review, synthetic media on social platforms has seen exponential growth, with over 1.5 billion views of identified synthetic tweets in less than a year, highlighting the urgent need to understand and address the implications of this evolving technology.

Types of Deepfakes: Identifying the Dangers

murf.ai

Face-Swapping - It's the most commonly known deepfake, wherein someone's face is placed on another person's body. Depending on the quality, this can be realistic and may fool people into believing that the individual whose face has been used is the one in the manipulated photo or video.

Voice Synthesis - Even audio can be replicated. This technology copies the voice, intonation, and speech patterns of a person, allowing the tool to speak as if it's actually them. This can be deceiving when the victim's voice is used for fraud calls and recordings.

Motion Manipulation - Deepfake can alter a person's expressions in photos and adjust their movements on video, pretending as if the individual is reacting or moving in certain ways that did not happen in reality.

Fake Text Generation - Artificial intelligence can copy someone's writing style, voice, and tone, so it can also compose articles or statements as if they're the person in question. This can be used to spread misinformation and propaganda, especially if AI is impersonating a renowned or influential personality.

Object Alterations - Deepfake can also be in the form of objects getting removed, replaced, edited, or added to put the subject in a bad light. This is usually done to celebrities and politicians in order to smear their image.

6 sources

How Deepfake Works: The Technical Process of Switching Faces

towardsdatascience.c...

It all starts with a deep neural network. You must first feed it hours and hours of videos to train its algorithm to identify the movements and features of a human being

. This kind of machine learning gives the system a 3D visualization of a person that it can use as the foundation for deepfakes

. The next step is merging this trained neural network with computer graphic software. These two together can overlay the AI-produced motions and patterns onto actual videos of people, enabling them to change the facial and audio attributes of the human subjects

. This process seems complex, but it's actually not that hard as long as you have a basic grasp of graphic design and software development. Of course, you need to have more skills for more detailed and realistic deepfakes, but it's alarming how easy it is to create one as long as you have a video or an audio recording of a person

. With the undying popularity of social media platforms and the constant posting of users on them, it's even easier to collect such data from a specific person who is active online

5 sources

How to Recognize a Deepfake

glair.ai

Deepfake isn't perfect. There are signs that will tell you if a certain media is fake or not. All you have to do is be vigilant and look for the following indicators: Inconsistencies: Look for any discrepancies and quirks in the media like the person suddenly wavering on a video, the face not seeming to match the body in some photos, the voice warbling in a call, or if certain statements are uncharacteristic of a person.

Blurry Edges: Amateur hackers may not know how to refine deepfakes well so you may find blurry portions where the hair, face, neck, and body meet.

Unnatural Traits: Observe if the person in a video or image has an uncanny expression or skin that's either too smooth or too creased. If you're viewing a clip, watch the eyes and see if they blink at the same time or if they don't blink at all. The eye and hair color must also remain consistent from the beginning to the end of the video.

Inaccurate Lip Sync: Audio is placed over a video if you notice that the words don't match the mouth movements of a person when they're speaking.

Background Glitches: A manipulated video tends to have glitches in the backdrop, especially when the subject moves and the surroundings bend or morph.

Source Reputation: Check where the media came from or who posted it. If it's a dummy account or a shady website, the media may be a deepfake.

Deepfake Detectors: If you still feel suspicious of what you see or hear even if there's no stark sign that it might be a deepfake, you should run it through a deepfake detector just to be sure.

2 sources

How to Keep Yourself Safe From Deepfakes

Be mindful of what you post online: Sharing photos and videos on social media is already a norm, but this poses great risk for deepfakes. Be aware that everything you publish on the Internet may be used against you, so only upload media that you deem to be appropriate and safe. Limit your circle: To prevent strangers and hackers from grabbing your data, only allow certain people to view your social media accounts and pages. Adjust your privacy settings and only add or accept people that you personally know. Place watermarks: If you're an online creator or influencer, it's important to overlay a watermark on your posts. If a watermark is in the way, it may be harder for neural networks to process your media and hackers will be more likely to back off since your content is traceable. Fortify your accounts: If someone with ill intentions can't penetrate your data, they may try to take over your account instead to gain access to your posts and information. To deter these hackers, ensure your password is strong with a mix of letters, numbers, and other characters. You should also enable the feature of multi-factor authentication so you have double protection in case your first layer of security gets compromised.

Closing Thoughts on Deepfakes

With artificial intelligence getting smarter by the day, hackers have found methods to leverage the technology to assist in their crimes and scams. Deepfake wasn't deepfake when it was first introduced to the public. Before, it used to be a tool utilized in film to swap actors' faces with their doubles. The early apps with this feature were also used to make fun, silly images and clips where friends and family can switch appearances or ordinary people can pretend to be their celebrity idols.

Unfortunately, some people just have bad intentions. What was once a promising innovation became what everyone now recognizes as deepfake. The good outputs were forgotten because they were replaced with malicious impersonations, revenge porn, financial fraud, online scams, and fake news.

Thankfully, a lot of governments and huge social media companies have set laws and regulations against the creation and spreading of deepfakes. However, these aren't absolute and there are still plenty of posers hiding behind their screens, just waiting to steal your data and create deepfakes out of them. It's your responsibility to secure your information and lower the possibility of being their next victim.