Identifying Deepfakes

May 15, 2024 by
Mark Nash

Have you noticed videos recently of public figures saying outrageous things? Well, you should probably double-check whether they are real videos or if they are completely fabricated. Welcome to the often terrifying world of deepfakes, a rapidly evolving technology that uses artificial intelligence (AI) to create synthetic media, including videos or audio recordings. While deepfakes can be used for creative purposes, such as satire or entertainment, their potential for misuse is something that needs to be addressed. To see an example, you need look no further than the fact that deepfakes are already being used in political campaigns to spread disinformation. During the 2024 election season, a fake robocall used deepfake technology to mimic the voice of a candidate to fool people into believing they said something they never said. Bad actors are using deepfakes to create phishing attacks, spread misinformation, damage reputations, and even manipulate financial markets. Being able to identify different types of deepfakes is now a crucial skill to stay safe and well-informed. So, how many types of deepfakes are there, and how can you spot them?

Face-Swapping Deepfakes

Face-swapping deepfakes are where one person's face is superimposed onto a different person in an image or video. These can be shockingly convincing, especially when using high-quality footage and advanced AI algorithms. These deepfakes are used by scammers to spread misinformation and sow distrust. Here's what to look for to identify these deepfakes:

  • Look for glitches/inconsistencies:¬†Pay close attention to lighting, skin tones, and facial expressions; if they appear unnatural or inconsistent throughout the video, that could be a sign of being a deepfake.¬†Look for subtle glitches, such as hair not moving realistically or slight¬†misalignments around the face and neck.
  • Check the source:¬†Where you find the video can be a good indicator of the likelihood of it being a deepfake. Videos are more likely¬†to be¬†legitimate if they are from reputable news sites rather than a random social media page. Be cautious of unverified sources and unknown channels.
  • Listen closely:¬†Does the voice sound natural? Does it match the person's typical speech patterns? Incongruences in voice tone, pitch, or accent can be giveaways.

Deepfake Audio

This type of deepfake involves creating synthetic voice recordings that mimic a specific person's voice by copying their speech patterns and intonations. Scammers use these deepfakes to create fake audio messages to make it seem like someone said something they didn't. These deepfakes are seeing a considerable rise in use for phone-based scams and phishing attacks. Here's what to look for to identify these deepfakes:

  • Pay attention to audio quality:¬†Deepfake audio can sound slightly robotic or unnatural¬†when¬†compared to the genuine voice of the person they are attempting to mimic. Listen for pauses or emphasis in strange places or inconsistent pronunciation.
  • Consider what is said:¬†Consider if what was said seems like something the person would¬†actually¬†say or if it matches the context in which it's presented. If the content seems out of character or contradicts known facts, that could be a sign it is a deepfake.
  • Seek verification:¬†Similar to how you would vet a potential phishing attack, see if you can externally verify if the audio does come from who it seems and if the things it says are true. If you cannot verify its authenticity, approach it with healthy skepticism.

Text-Based Deepfakes

A type of deepfake seeing a dramatic rise in usage is AI-generated written content, such as social media posts, articles, or emails. While AI-generated text is used regularly by many, text-based deepfakes specifically attempt to mimic the writing style of a specific person or publication. As with any type of deepfake, scammers use them to spread misinformation and impersonate others online. However, these deepfakes are also being used a lot by hackers to create phishing attacks. Here's what to look for to identify these deepfakes:

  • Read critically:¬†Pay attention to¬†the writing style, vocabulary, and tone. Does it match the way the person or publication typically writes? Look for unusual phrasing, grammatical errors, or inconsistencies in tone.
  • Fact check:¬†Don't unquestioningly trust things you read online, especially from just one source. Check reliable sources to verify the accuracy of the information presented in the text.
  • Identify emotional triggers:¬†Learn to identify content that is trying to evoke visceral emotions, such as fear, anger, or outrage. Scammers frequently use these techniques to cloud your judgment and get you to fall for scams.

Deepfake Videos with Object Manipulation

This type can be the hardest to spot, as it often tries the hardest to look legitimate. Bad actors use AI to manipulate objects in video footage to fabricate events or alter visual evidence. Here's what to look for to identify these deepfakes:

  • Scrutinize the movement:¬†Pay attention to how objects move in the video and if it seems natural and consistent with the laws of physics. Unnatural movement patterns, sudden changes in the size of objects, or inconsistencies in lighting and shadows can all be signs of a deepfake.
  • Seek original footage:¬†If you can find the¬†original¬†source of the video footage,¬†then¬†you can directly compare it to the manipulated version and identify alterations.

Staying vigilant and applying critical thinking are crucial in the age of deepfakes. Familiarize yourself with the different types of deepfakes and learn how to recognize potential red flags. Before you believe something outrageous, take some time to verify the information through reliable sources. Taking these kinds of steps will help you become more informed and stay protected.