Injecting imperceptible noise to disrupt AI synthesis… advancing the technology further

Seoul — Lavarwave, a company specializing in digital crime response, announced on the 2nd that it will provide a public hands-on experience of a deepfake pre-prevention technology developed jointly with the KAIST Cybersecurity Research Center, in collaboration with the Korean National Police Agency.

As deepfake crimes rapidly increase, this initiative is drawing attention as the first case of directly introducing a fundamental technological prevention method to the public, moving beyond the limitations of post-incident response.

The newly revealed deepfake pre-prevention technology is based on adversarial attack techniques.

By injecting imperceptible micro-noise into a person’s photo, the technology causes severe distortion when AI attempts to use the image for deepfake generation.

This approach preserves the visual quality of the original image while technically suppressing the creation of deepfakes.

Through its collaboration with the National Police Agency, Lavarwave aims to allow anyone to directly experience a feature that protects their photos from deepfakes.

The nationwide experience service will be accessible through the National Police Agency’s Cybercrime Reporting System as well as Lavarwave’s website, blog, and Instagram to maximize accessibility.

In line with the rapid advancement of deepfake generation AI, the company is also pursuing continuous technological upgrades. In particular, it is conducting research to respond to the latest deepfake generation models and OpenAI models equipped with preprocessing removal capabilities. The strategy aims to achieve two goals simultaneously: suppressing the spread of criminal content and promoting positive innovation in response to the misuse of technology.

Lavarwave CEO Kim Jun-yeop said, “Starting with this experience service, our ultimate goal is to create an environment where anyone can easily protect themselves from deepfakes.”