Detecting deepfakes vital for a trustworthy digital future
This article is authored by Nimeshkumar Patel, senior network engineer and architect, Humana Inc.
Living in the contemporary world means information can be created, spread and shared with little effort. However, as new platforms of media enter into the social realm they are accompanied by employments which tend to deceive the masses and erode their credibility. Among the most concerning technologies in this realm are deepfakes: Images, voices, and videos depicting fake persons and situations which seem real and can be closely associated with the real actions of a person. With the worry of lots of deepfake content flooding the marketplace, it’s high time to build and establish deepfake detection systems.
Deepfakes are made with the help of Artificial Intelligence, and based on deep learning, using generative adversarial networks (GANs). In other words, a GAN is a contest between two algorithms: a generator that generates the fakes and a discriminator that seeks to identify them. For thousands of cycles, the generator optimises its content creation results and it becomes hard for the discriminator to distinguish a fake from a real data sample.
This separation is something that the governments, companies, and media organisations all need to get right to restore the credibility they have among the citizens. We can look at the stakes in the world of journalism. In the age of ‘fake news’ adding actual deep fakes into the mix can only decrease the credibility of the source. Consider a news channel putting out live a fake video of a political figure, supporting toxic policies.
Often the later creation of a follow-up video debunking the deepfake could be helpful but, in many cases, the damage to the reputation of the targeted politician is already done.
However, it is important to know that deepfakes are a threat to democracy. They remain so influential that they can twist the outcome of any given election by promoting fake news at an unprecedented speed. In the election period, deepfakes would create misunderstandings and propaganda among voters, and spread false information about a particular candidate. When such deepfakes go unaddressed, individuals can no longer rely on any given video or audio as genuine, which means that the population will lose faith in the very media that are expected to steer democracies forward. This disintegration of trust could lead to the creation of cynicism and political indifference destroying the democratic framework of knowledge-based decision-making.
Another likely application of deepfake technology emerged as one of the worst use cases: Revenge pornography where victims’ faces were then grafted onto pornographic content they made with someone else. This invasion of privacy not only victimises people but may cause individual psychological trauma, social exclusion, and reputational damage.
In dealing with the problems of deepfakes, one has to proceed with the choice of an interdisciplinary approach that would encompass many strands of work. To start with, there is a need to invest in detection technology. Advertisers, customers, and governments have to contribute with resources to fund ongoing research that will ensure that detection tools remain as effective as possible in identifying these deceptive tactics. More responsibility also lies with social networks.
The awareness of the public is also important. It means that by raising awareness of the general public in regards to deepfakes and independent models’ training in detecting fake content, it will be easy to curb fake content distribution. This education could spread to schools so that learners develop relevant skills as they encounter a challenging media environment. At the same time, media organisations require new means to address the issue of deepfake verification as fact-checking procedures should be adapted to such new phenomena. Through this, they can also increase their credibility by publicly airing their fake detection programmes.
We are at a stage where technology can either build on that trust or start taking it apart. It’s time to take deepfakes seriously and protect the truth in the internet space to maintain its key value today. Older generations believe that trust is the basis of society and without it, people run around spreading rumours, fear, and doubts. Never was the protection of that trust more important than it is today, and the ability to counter deepfakes is one of the most important weapons in that fight.
This article is authored by Nimeshkumar Patel, senior network engineer and architect, Humana Inc.