Hancom With announced Friday that its deepfake detection and speaker authentication solution, 'Speeki,' achieved notable results at an international conference and competition hosted by the International Speech Communication Association.
The AI Security Research Center at Soongsil University, which conducts research on Speeki's core technologies, had seven papers accepted at this year's International Conference on Spoken Language Processing (INTERSPEECH). The papers covered research related to speech recognition, speaker verification, and detection of synthetic voices.
The center also secured second place in the 'WildSpoof Challenge' organized by the Acoustics, Speech, and Signal Processing Society under the Institute of Electrical and Electronics Engineers (IEEE). This competition evaluates voice deepfake generation and detection performance based on data collected from everyday scenarios.
Notably, in the 'SpoofCeleb' evaluation—which reflects real-life background noise and diverse recording environments—the team achieved higher performance than the top-ranked overall team. This demonstrates their technical capability to identify synthetic voices even in environments mixed with call noises or ambient sounds.
Based on these research achievements, Hancom With plans to enhance Speeki's AI synthetic voice detection accuracy and expand its business scope into financial sector voice authentication and call center voice phishing detection.
Song Sang-yeop Han Keom-wi-deu (CEO) stated, "These results demonstrate that Speeki's technical reliability and differentiation are on a global level," adding, "We will target the next-generation security market to combat cyber crimes exploiting generative AI."
in TECH
