South Korean Police Unveil Deepfake Detection Tool to Combat AI-Generated Video Crimes | Be Korea-savvy

South Korean Police Unveil Deepfake Detection Tool to Combat AI-Generated Video Crimes


This newly developed software by the police can analyze videos suspected of using deepfake technologies like Face Swap within 5-10 minutes and determine their authenticity. (Image courtesy of Yonhap)

This newly developed software by the police can analyze videos suspected of using deepfake technologies like Face Swap within 5-10 minutes and determine their authenticity. (Image courtesy of Yonhap)

SEOUL, Mar. 6 (Korea Bizwire) – The National Police Agency’s National Office of Investigation has announced the development of a deepfake detection software, marking a significant step in the crackdown on crimes associated with deepfake technology.

Deepfakes, a blend of ‘deep learning’ and ‘fake,’ involve overlaying existing videos with other images using AI to create fake media. 

This newly developed software by the police can analyze videos suspected of using deepfake technologies like Face Swap within 5-10 minutes and determine their authenticity.

Upon completion, the software generates a report instantly, aiding in immediate investigative use.

Previously, deepfake detection models were primarily developed overseas, focusing on Western data, leading to lower detection rates for videos featuring Koreans.

The new software, however, is trained on over 5.2 million data points, including 1 million related to Koreans and 130,000 of Asian ethnicity, featuring 5,400 individuals.

With the application of the latest AI models, the software can adapt to newly emerging synthetic video forms, ensuring robust responsiveness.

The detection accuracy stands at approximately 80%. Nevertheless, the police plan to use the software more as a guide for investigative direction rather than direct evidence, considering the detection rate isn’t perfect.

Given the continuous evolution of deepfake technology, the police aim to minimize false positives by undergoing cross-verification with academia and industry AI experts for deepfake-related electoral crimes. 

An official from the National Office of Investigation expressed confidence in the software’s ability to facilitate proactive investigations by enabling swift analysis and result verification when deepfake usage is suspected.

Efforts will continue to refine the software for higher accuracy and extend its application beyond electoral and synthetic exploitation crimes to various deepfake-related offenses.

The increasing misuse of deepfakes globally, especially in distorting public opinion through fake news, raises concerns domestically, especially with the upcoming general election.

The use of deepfakes in election campaigns is strictly prohibited during the election period, and the police have committed to a strict response strategy, with cyber investigation units taking charge of such electoral crimes.

Kevin Lee (kevinlee@koreabizwire.com) 

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>