What is a deepfake and how to avoid it

While artificial intelligence has created new conveniences in many areas of daily life and technology, its misuse has also given rise to a threat called ‘deep fake’.

Fake images, videos and audios created with the help of AI are so close to reality that it is sometimes extremely difficult for an ordinary user to distinguish between real and fake content.

Through deep fake technology, artificial content can be created that appears to be completely real by imitating a person’s face, voice, movements and speech patterns. For this purpose, generative AI and deep learning technologies are usually used, which create new content by learning different patterns from available images, videos and audio samples.

Deep fake is not limited to video

Deep fake is usually mentioned in relation to fake videos, however, voice can also be cloned to some extent through artificial intelligence. This is called ‘voice cloning’ or ‘deep fake audio’. AI can create a highly similar voice by understanding a person’s tone, pronunciation, and speaking style from a short sample of their voice.

This capability can also be used for cyber fraud. Fraudsters can try to obtain money, passwords, OTPs, or other sensitive information from a person by imitating the voice of a relative, friend, officer, or even a bank representative.

How to recognize a fake video?

A few unusual signs can alert a user when watching a deepfake video. For example, if the lip movement and voice in the video do not match each other or the facial expressions seem unnatural, the video may be suspicious. Similarly, eye movement, focus, and blinking patterns may also appear different from normal human behavior.

If the head and neck movements seem disconnected from the rest of the body, the movements of the hands and fingers are unusual, or the edges of the face appear blurred, it is also important to check the authenticity of the video. In some fake videos, the skin looks overly clear or artificial, while the lighting and shadows do not match the face or body in a natural way.

Audio can also help identify fake content. Unnatural fluctuations, robotic voices, or sudden changes in voice can be signs that the audio has been artificially altered.

Three-finger roll on video calls

Some experts also mention a method like the ‘three-finger roll’ to quickly check a suspicious person during a video call. In this, the person on the video call is asked to hold up three fingers in front of his face.

Some deepfake or face swap technologies are not able to process such sudden movements correctly, which can result in visual distortions around the face or fingers. However, this method is not definitive or 100% reliable proof, as modern AI systems can sometimes successfully pass such tests. Therefore, it should only be used as a preliminary check.

How to avoid deepfake fraud?

It is important to verify any suspicious video, audio, phone call or link before immediately believing it. You should also avoid opening unverified links received through social media and messaging apps.

If someone demands money, password, OTP or any verification code over a phone or video call, be sure to check their identity with another trusted source. Personal information, banking details, passwords or other sensitive information should not be shared with a suspicious person.

You need to be especially careful if the voice or appearance of the person in front of you seems completely familiar, because the main purpose of deepfake technology is to create a perception of reality.

What to do if you fall victim to deepfake fraud?

If someone falls victim to fraud through AI or deepfake, instead of panicking, you should take necessary steps immediately. Registering a complaint by contacting the relevant cybercrime authorities or the police is an important part of this process.

All relevant evidence, including fake video or audio, screenshots, links, phone numbers, messages and chats, should be preserved to aid in the investigation. If the fake content is present on a social media platform or website, it should also be reported through the relevant platform’s reporting system.

The use of artificial intelligence is likely to become more common in the coming time, so preventing deepfakes is not just a technology issue but also a requirement of digital awareness. Fact-checking any viral video or sensational audio before sharing it further is the most effective way to protect the user from fraud, misinformation and potential financial or social harm.