Today, the government emphasized the growing danger posed by deepfakes to democratic processes and revealed plans to introduce new regulations to tackle this issue.
Ashwini Vaishnaw, the IT Minister, stated that social media platforms recognize the importance of implementing practical and impactful measures in key areas like detecting and preventing deepfakes, improving reporting systems, and educating users.
What Is Deepfake?
A deepfake refers to a kind of fake content made using artificial intelligence to alter or create visual and audio material. Typically, these alterations are done with harmful intentions to make them seem real.
MIT mentions that the term “deepfake” became widely known in late 2017. It originated when a Reddit user named “deepfake” established a platform on an online news site. This platform was used to distribute adult videos created through open-source face-swapping technology. Deepfake employs a type of AI known as deep learning to craft images or videos portraying fictitious events.
How Deepfakes Are Threat?
Deepfakes, a concerning mix of authentic and manipulated media, pose a significant threat to public trust and reality. These manipulative creations, using both video and audio elements, have the power to distort public perception, spread false information, and harm reputations by making people appear to say or do things they never did.
In the wrong hands, like those of cybercriminals, deepfakes transform into potent weapons capable of disrupting and even dismantling businesses and governments. For instance, a fake video featuring a company’s top executive or a key political figure can severely damage the reputation of the entity involved.
Instances of deepfake videos going viral on social media have become increasingly common. A recent case involved a deepfake video targeting the actor Rashmika Mananna, raising serious concerns about the misuse of this technology. In response, the Indian government is taking steps to address and counter this growing threat.
Creating deepfakes typically involves employing deep neural networks and a technique known as face-swapping. The process relies on a base video, which serves as the foundation for the deepfake, and a compilation of video clips featuring the target individual.
The rapid advancement of artificial intelligence has also led to a disturbing surge in deepfake pornography. This involves the production of hyperrealistic images and videos with minimal effort and cost, further exacerbating the challenges posed by this technology.
How To Spot A Deepfake
According to MIT, while there’s no foolproof way to completely protect against deepfakes, there are certain signs that can help you verify the authenticity of online content.
In the widely shared Rashmika Mandana deepfake video, many viewers missed the initial difference in the person’s face. The video featured Ms. Mandanna’s face placed onto the body of Zara Patel, a British model and Indian-origin influencer who originally created the video. The deepfake only became apparent when the person fully appeared on the screen.
In deepfakes, there may be issues with lip movements or blinking not matching seamlessly. This happens because AI algorithms might struggle to precisely track eye and mouth movements. However, as AI continues to improve, distinguishing between fake and real content becomes more challenging with the advanced tools available.
Highlighting the shared responsibility for deepfake content, IT Minister Mr. Vaishnaw suggested potential penalties for both creators and hosting platforms. The government is considering measures to hold both parties accountable.
To effectively combat the spread of deepfakes, a comprehensive approach involving regulations from government agencies and collaboration with technology and social media companies is essential. Developing cross-platform detection tools through alliances is crucial in addressing this issue.
How Are Deepfakes Made?
Creating deepfakes involves various techniques. One widely used approach is the application of generative adversarial networks (GAN). These networks teach themselves to identify patterns through algorithms, allowing the generation of fake images.
Another method involves AI algorithms known as encoders, specifically utilized in face-replacement and face-swapping technology. In this process, the decoder retrieves and exchanges facial images, allowing one face to seamlessly blend with an entirely different body.
Deepfakes utilize autoencoders, which extend beyond the capabilities of traditional encoders by enabling cybercriminals to craft entirely novel images. Deepfake applications leverage two autoencoders, facilitating the transfer of images and movements from one source to another.
Also Read: Monitoring Delhi Air Pollution and Weather: Apps for AQI Data Tracking
What Are Deepfakes Used For?
Deepfake technology, a concerning innovation, has the potential for various malicious applications, including:
Deceptive Schemes and Falsehoods:
Cybercriminals leverage deepfake technology to fabricate scams, false assertions, and hoaxes, undermining and destabilizing organizations. For instance, an attacker might produce a deceptive video featuring a senior executive confessing to criminal activities, imposing a significant impact on the business’s brand, public image, and stock value.
Unauthorized Use of Celebrities:
Nonconsensual pornography, constituting a substantial portion (up to 96%) of internet deepfakes, predominantly targets celebrities. Deepfake technology is exploited to create deceitful instances of revenge porn, posing a significant threat to personal privacy.
Manipulation of Elections:
Deepfake videos have been employed to circulate fabricated content featuring world leaders like Donald Trump and Barack Obama, raising concerns about their potential use in election manipulation. The 2020 U.S. election campaign, for instance, faced widespread apprehensions regarding the impact of deepfake videos.
Social Engineering Exploits:
Social engineering scams incorporate audio deepfakes, deceiving individuals by making them believe trusted figures have uttered statements they never made.
An illustrative incident involved a U.K. energy firm’s CEO who fell victim to a deepfake voice impersonating the chief executive of the parent company, resulting in a €220,000 transfer to a purported Hungarian supplier.
Automated Disinformation Campaigns:
Deepfake technology facilitates the dissemination of automated disinformation attacks, promoting conspiracy theories and incorrect narratives on political and social issues.
Notably, a fabricated video of Facebook founder Mark Zuckerberg falsely claiming to control billions of people’s data through the fictional organization Spectre, from James Bond lore, exemplifies this misuse.
Identity Theft and Financial Deception:
Deepfake technology is exploited for identity theft and financial fraud, allowing perpetrators to create new identities or assume those of real individuals. Attackers use the technology to forge documents or mimic their victim’s voice, enabling them to establish accounts or make purchases while impersonating the targeted person.
Deepfake FAQs
Q.1. What is a Deepfake?
Ans. A deepfake is fake content created using artificial intelligence to manipulate visual and audio material. Typically crafted with harmful intent, deepfakes aim to deceive by making altered content appear authentic.
Q.2. Why are Deepfakes a Threat?
Ans. Deepfakes, blending authentic and manipulated media, pose a threat to public trust. They can distort perceptions, spread false information, and harm reputations by portraying individuals saying or doing things they never did. In the wrong hands, deepfakes become potent weapons, capable of disrupting businesses and governments.
Q.3. How Can You Spot a Deepfake?
Ans. While not foolproof, signs of deepfakes include discrepancies in facial features, lip movements, and blinking. As AI advances, distinguishing between real and fake content becomes more challenging. Shared responsibility is crucial, and potential penalties for creators and hosting platforms are being considered.
Q.4. How are Deepfakes Made?
Ans. Deepfakes use generative adversarial networks (GAN) and AI algorithms like encoders for face-replacement and face-swapping. Autoencoders, extending beyond traditional encoders, enable the creation of entirely novel images, facilitating the transfer of images and movements from one source to another.
Q.5. What are Deepfakes Used For?
Ans. Deepfakes have malicious applications, including fabricating scams, unauthorized use of celebrities for revenge porn, manipulation of elections through fake videos, social engineering scams using audio deepfakes, automated disinformation campaigns, and identity theft for financial fraud. They pose serious risks to individuals, businesses, and governments.
