Deepfake Voice Technology: A Mimicry or Manipulation in 2021?
Controversy regarding whether deepfake voice technology is a mimicry or manipulation in 2021.
Are you confused about whether to believe the content of any celebrity video or not in this 21st century? Yes, we all are! The credit goes to the popularity of AI that led to the advent of the deepfake technology. At first, it was used for fun, mimicry, medical and other entertainment purposes but currently, it gave rise to manipulation from dark web hackers that can create havoc in the victim’s life. There is a drastic shift from acquiring entertainment to experiencing harassment in this tech-driven era due to the rise in deepfake voice technology. Let’s dig into the minute details on this deepfake voice technology.
How does deepfake voice technology work?
The updated computer graphics and AI machine learning algorithms provide voice cloning of celebrities to create content that never happened before. Producing deepfakes is much easier and cost-effective for any creator who acquires the knowledge of the neural network. GAN (Generative Adversarial Networks), an AI algorithm, acts as a primary engine of developing deepfake voice technology. But nowadays, hackers are not using GAN to manipulate scams with deepfake voice technology. Hackers tweak AI algorithms by using spyware to collect voice recordings of their potential victim for successful manipulation. The data sources can be from uploaded interviews, social media videos, phone calls, mobile voicemails or regular conversations. After collecting the raw data of voice modulations, the AI starts working on mimicking the victim with NLP. It is time-consuming for AI to develop the exact voice with tones, accents and volume. The final audio deepfake files are both impressive and intimidating at the same time with utmost high-quality audio having similarities with the original voice.
Voice phishers are one of the potential threats for the society due to the ability of malicious hackers to circumvent the corporate VPN. The hackers can initiate blackmailing by integrating deepfake voice technology in some videos with explicit sexual or violent scenes that can go viral on social media. But there is a restriction in the usage of these AI technologies to protect the privacy of society.
There are various companies launching new AI algorithms to create an advanced deepfake voice technology such as the Baidu Deep Voice Research team has launched a ground-breaking technology that clone voices within 30 minutes from providing sufficient data, Adobe has launched a programme that can mimic voices within 20 minutes and so on. These are launched for the positive sides of deepfake voice technology such as providing a voice to the patients in need like deaf, Parkinson’s disease or ALS, dubbing an animated movie, educational purposes, audiobooks for children and so on. These all provide high quality listening experiences that enhance the standard of living worldwide.
How will you detect deepfake voice technology?
Since there is a massive rise in fraudulent activities with voice cloning and the cons are more than the pros, there is always an urge to innovate smarter technologies to combat the harassments from manipulations by those malicious hackers. Voice anti-spoofing is the AI technology that can distinguish between live voice and the manipulated voice effectively and efficiently. It was initially launched to solve the voice biometric spoofing but it shifted the focus to detecting the artefacts absent in the live voice.