Deepfake-as-a-Service Now Offered on the Darknet
Specialists at Kaspersky Lab have uncovered advertisements on the darknet promoting real-time video and audio deepfake services. Prices vary by complexity and duration, starting at $50 for video and $30 for voice deepfakes.
From costly experiments to cheap services
Offers to create deepfakes on the darknet are not new, but the latest findings highlight a major shift: real-time generation of fake audio and video, at far lower prices. In 2023, a single minute of deepfake video could cost up to $20,000. Today, entire sessions can be purchased for a fraction of that.
How the services work
The advertisements promise several options, including:
- Real-time face replacement during communication via video conferencing platforms or messengers.
- Face substitution for passing identity verification checks.
- Camera feed manipulation, replacing images from a phone camera or virtual device.
Vendors also claim to offer supporting tools such as:
- Lip-sync software capable of matching any text to a subject’s mouth movements, even in foreign languages.
- Voice-cloning utilities that alter tone and timbre to express specific emotions.
Caveats and risks
Kaspersky researchers caution that some of these advertisements may themselves be scams designed to defraud potential buyers, rather than deliver working software.
Expert perspective
“On shadow marketplaces, we see not only advertisements for ‘deepfake-as-a-service’ but also clear demand,” said Igor Kuznetsov, Director of Kaspersky’s Global Research and Analysis Team (GReAT). “Malicious actors are actively experimenting with AI services. We’re also observing more advanced tools on the darknet—malicious large language models created from scratch, running locally without relying on mainstream LLMs.
“Although these technologies do not create fundamentally new cyber threats, they significantly expand attackers’ capabilities. In such conditions, cybersecurity professionals must work harder to keep pace. Paradoxically, one of the best defenses may be to use AI ourselves—to boost the productivity of defenders and the effectiveness of protective solutions.”
