DistantNews
🇰🇷 South Korea /Technology

Soongsil University AI Safety Research Center develops voice cloning prevention technology 'RoCo', adopted by ICASSP 2026

From Hankyoreh · (11h ago) Korean Positive tone

Translated from Korean, summarized and contextualized by DistantNews.

TLDR

  • A South Korean research team has developed 'RoCo,' an AI technology designed to prevent voice cloning.
  • The technology, which applies protective processing to user voices, will be presented at the ICASSP 2026 conference.
  • RoCo aims to combat voice-based fraud and fake news by making cloned voices unusable for AI learning.

Researchers at Soongsil University's AI Safety Research Center have achieved a significant breakthrough in combating the rising tide of AI-driven voice manipulation. Led by Professor Choi Dae-sun, the team has developed an innovative AI voice-cloning prevention technology named 'RoCo' (Robust Code). This technology is set to be unveiled at the prestigious ICASSP 2026 conference in Barcelona, Spain, marking a notable achievement for Korean AI research on the global stage.

RoCo operates by applying a protective layer to an individual's voice during recording. While sounding natural to human ears, this processed audio is designed to prevent AI systems from accurately learning its characteristics. This approach directly addresses the growing threat of voice phishing, financial fraud, and the spread of fake news, which increasingly leverage sophisticated AI voice-cloning techniques. The development is particularly timely, given the escalating sophistication of deepfake and generative AI technologies.

RoCo is differentiated from existing methods by applying protective signals to the AI's voice recognition process, rather than altering the voice itself.

— Choi Dae-sunSoongsil University professor explaining the unique mechanism of the RoCo technology.

What sets RoCo apart from previous methods is its novel insertion of protective signals within the AI's voice analysis process, rather than merely adding imperceptible noise. This method ensures the technology's effectiveness even against advanced noise-reduction techniques and significantly reduces processing time, making it approximately five to ten times faster than existing solutions. Professor Choi emphasizes that RoCo's strength lies in its ability to prevent the generation of fake voices rather than merely detecting them after the fact. This proactive stance, developed within South Korea's burgeoning AI research landscape, offers a crucial defense mechanism against a rapidly evolving digital threat, as highlighted by the Hankyoreh's reporting on this innovative Korean technology.

The goal was to implement a defense that cannot be erased.

— Kim Seung-minStudent researcher Kim Seung-min discussing the objective of developing an unerasable defense technology.
DistantNews Editorial

Originally published by Hankyoreh in Korean. Translated, summarized, and contextualized by our editorial team with added local perspective. Read our editorial standards.

More Stories

From Our Blog