ETRI Filters Out 99% of Harmful Videos Using AI

[Asia Economy Reporter Eunmo Koo] The Electronics and Telecommunications Research Institute (ETRI) announced on the 9th that it has developed a technology that analyzes the harmfulness of video content using artificial intelligence (AI) technology.

The AI searches web pages using registered keywords and analyzes posts such as text and images on the web pages to detect harmful content.

The accuracy of harmfulness judgment reaches 99.4%, and the detection time is also fast, taking less than 0.01 seconds, demonstrating excellent performance.

During a two-month pilot operation period starting in June, a total of 18,945 websites were automatically collected using video search keywords, thumbnail images, URLs, and more. Among them, 2,631 websites were successfully filtered out as harmful sites.

Namkyung Lee, head of the Media Intelligence Research Division at ETRI, said, “This AI technology will be usefully applied to support victims of digital sexual crimes,” adding, “We will continue to support minimizing the distribution and spread of illegal recordings.”