top of page
  • Raghav Sand

Deepfake: The Dark Side of Internet

The internet has democratized communication and access to information. In the past two decades nothing has matched the power and popularity of internet. Ever-increasing stream of content is being uploaded on websites and social media platforms. Among all the societal benefits of the internet, there is a dark side to it. The cat and mouse game between the regulators and offenders is currently being won by the latter. The internet was conceived as a level playing field with little or no restriction. Whosoever wishes to use this platform to propagate falsehood, is in essence trying to defeat the purpose of human ingenuity.

What is a Deepfake?

As per the Cambridge English Dictionary, a deepfake is a video/image or sound recording that replaces someone’s face or voice with that of someone else, in a way that it appears real. The primary objective of applying deepfake is to spread misinformation. Advancement in technology has blurred the line between real and fake. Persuasive counterfeit content is posing a threat to individual privacy and democratic governments. Mimicking speech pattern and appearance may seem like fun and games, but the consequences of a deliberate misuse of an individual’s biometrics can be devastating.

Everyone with a smartphone has either received or created superimposed pictures of themselves or their friends and relatives over a celebrity. A popular app called ‘FaceApp’ lets users see an older version of themselves or any other person’s image they wish to tamper with. These unthoughtful actions may dilute privacy and can lead to financial or emotional turmoil in the years to come.

Seeing is Believing (Not Anymore)

People have overturned conventional wisdom as per their wishes. The modern-day equivalent of the phrase ‘seeing is believing’ is that people see what they want to believe. A divided world with disregard for other person’s views has become a boon for the architects of chaos. Public opinion is important for common people as much as it is for politicians, sportsmen and artists.

Even without verifying the authenticity of a photograph or video, most individuals form an indelible opinion. By the time a fact-checking exercise is concluded, the masses have already moved on to the next trending content. Once again, the purpose and reach of internet goes in vain.

Image Courtesy: Facebook

Using Artificial Intelligence to Detect Deepfake

The most fruitful approach in detecting a deepfake is to look for irregularities. Analysis of mouth formations (visemes) and phonetic sounds (phonemes) are being applied in both manual and mechanical detection procedures. Researchers look at the person’s mouth when making the sounds of a “B,” “M,” or “P,” because it’s almost impossible to make those sounds without firmly closing the lips.

Initially researchers applied manual technique, in which human observers considered frames of video. The results were encouraging, but the whole process was time-consuming and laborious. Subsequently, the researchers then tested an AI-based neural network, which was much faster, to make the same analysis after training it on videos.

Four to five Indian states have assembly elections scheduled in every calendar year, with the general elections scheduled every five years. Peripheral organisations are sometimes employed by the information technology / social media cell of political parties to do the dirty business. An even bigger threat is the interference by a foreign country. The United States Presidential election in 2016 and a doctored clip of President Joe Biden during his campaign trail are most apparent and striking examples.

Common citizens may not have sophisticated software and hardware to detect a deepfake or even a shallowfake, but if one pays attention a large chunk of the problem can be eradicated. Once users get better at detecting a fake, they will eventually discard the creators or platforms that publish deepfake content.

Social media companies thrive on user engagement. Selective and passive policing will not serve any purpose. What if the big tech companies come together or in their individual capacities develop and deploy techniques for detecting deepfake? Any and every move a commercial venture makes revolves around metrices and spreadsheets – and expecting a solution from social media giants is unrealistic optimism. We are on our own – the sooner you realise it, the better.

bottom of page