In the current era of digital advancements, discerning between authenticity and deception has become an arduous task. The emergence of deepfakes has further exacerbated this predicament, transcending the boundaries of the virtual realm and infiltrating our physical reality. Deepfakes encompass fabricated videos or images that employ artificial intelligence to generate convincingly counterfeit content. This technological innovation possesses the capacity to inflict damage and upheaval upon society through diverse means. On November 07, 2023, Rashmika Mandanna, an actor, vehemently denounced the authenticity of a scandalous video that had emerged on various social media platforms. The video featured her face, unknowingly superimposed onto the body of a British Indian influencer. This incident serves as a prime example of a 'Deepfake' video, where Artificial Intelligence tools were employed to manipulate images and videos.
What is deep fake?
Deepfake, an AI-driven technology, is employed to create or modify video content in a manner that portrays events that never actually took place. This innovative technology operates by utilizing deep learning neural networks to manipulate both facial features in videos and vocal characteristics in audio recordings. These neural networks function in a similar fashion to the human brain's information processing capabilities. To train these networks, extensive datasets comprising thousands of images or voice samples are employed. The accessibility and affordability of deepfake technology have significantly increased, which identifies a growing prevalence of "cheap fakes." As any technology becomes more widely available, it inevitably attracts the attention of cybercriminals who exploit it for their illicit gains. This trend is observable not only in the realm of deepfake technology but also in other domains such as the Internet of Things (IoT), cloud computing, and artificial intelligence.
Process of creating Deepfakes
There are multiple techniques available for generating deepfakes. One widely utilized approach involves the implementation of a generative adversarial network (GAN). This network employs algorithms to train itself in recognizing patterns, and it can also be employed to fabricate counterfeit images. Another method involves the use of AI algorithms known as encoders, which are commonly utilized in face-replacement and face-swapping technologies. By utilizing a decoder, these encoders retrieve and interchange facial images, allowing for the superimposition of one face onto an entirely different body. Deepfakes employ autoencoders, which surpass the compression and decompression capabilities of traditional encoders, thereby enabling cyber criminals to create entirely novel images. Deepfake applications utilize two autoencoders, facilitating the transfer of images and movements from one image to another.
Several applications and software programs are available to assist novice beginners in creating deepfakes. These include the Chinese app Zao, DeepFace Lab, FaceApp (a photo editing app equipped with AI techniques), Face Swap, and the now-banned DeepNude, a highly dangerous app that generates fake nude images of women. GitHub, an open-source online software development community, hosts a significant number of DeepFake apps.
Identifying Deepfakes
Deepfakes can be identified through various indicators of unnatural activity or movement. One such indicator is unnatural eye movement, as deepfakes often struggle to replicate the natural way in which people's eyes follow and react to the person they are speaking with. Additionally, a lack of blinking is a common flaw in deep faked videos, as it is challenging to replicate the regular blinking action that occurs in real life. Another telltale sign is unnatural facial expressions and facial morphing. Deepfake technology involves stitching facial images from one source onto another, resulting in unusual or unnatural facial expressions. Furthermore, if a person's body appears to have an unnatural shape, it is likely a deepfake. Deepfake technology primarily focuses on manipulating faces rather than the entire body, leading to distorted body shapes. In terms of appearance, deepfakes often fail to generate realistic individual characteristics, such as frizzy or messy hair. They also struggle to replicate natural skin colors, resulting in abnormal skin tones. Awkward head and body positioning is another characteristic of deepfakes, with jerky movements and distorted images occurring when people move or turn their heads. Inconsistent facial positions are also common in deepfake images, further indicating their artificial nature. Deepfakes may exhibit odd lighting or discoloration, like unnatural skin tones, as well as misplaced shadows and unusual lighting. Lastly, deepfake videos often feature lip-syncing that does not align with the words being spoken by the individuals in the video. By being aware of these various indicators, individuals can better spot deepfakes and distinguish them from authentic content.
The legal stance on this matter according to our laws
Currently, India lacks a dedicated legislation addressing deep fake cybercrime; however, it can utilize a combination of existing laws to tackle this issue. If you or someone you know falls victim to a deep fake crime, these legal provisions can help. Section 66D of the IT Act stipulates “Whoever, by means of any communication device or computer resource cheats by personation, shall be punished with imprisonment of either description for a term which may extend to three years and shall also be liable to fine which may extend to one lakh rupees.” Section 66E of the IT Act stipulates “Whoever, intentionally or knowingly captures, publishes or transmits the image of a private area of any person without his or her consent, under circumstances violating the privacy of that person, shall be punished with imprisonment which may extend to three years or with fine not exceeding two lakh rupees, or with both.”
Deepfake cybercrime poses a significant threat to both individuals and organizations, as a single deepfake video can cause irreparable damage to individual reputation and brand. It is crucial for individuals and organizations to be aware of the risks and take proactive measures to protect themselves and their sensitive information. With the advancement of more sophisticated AI technology in the future, Deepfake may evolve further and pose more serious threats to the public. These threats could include election interference, political tensions, and an increase in various criminal activities.
The writer is Principal, Adarsh College, Rajdhanwar, Giridih, Jharkhand and Former Professor, Birla Institute of Technology, Mesra, Ranchi, Former Faculty Member BITS, Pilani. He has recently been granted his sixth patent by the Government of South Africa. The patent was awarded for his groundbreaking work in developing an algorithm that utilizes Artificial Intelligence to effectively combat the spread of epidemics.