Deepfake Detection
PALO ALTO, Calif., Aug. 20, 2024 – DeepBrain AI, a leader in generative AI technology, announced today the launch of a groundbreaking deepfake detection system developed in partnership with the Korean National Police Agency. This system is designed to combat the rising threat of deepfake crimes, which have become increasingly sophisticated and pose significant risks to society.
As concerns over deepfake technology grow, particularly in the context of phishing and election interference, this detection service allows users to analyze and verify the authenticity of digital content quickly and accurately.
“Our deepfake detection solution represents a significant leap forward in protecting society from the dangers posed by deepfake technology,” said Eric Jang, CEO of DeepBrain AI. “We will continue to advance the solution to prevent crimes that exploit increasingly sophisticated AI technologies to create social confusion and threats. This ongoing commitment includes enhancing our technology to stay ahead of emerging threats, ensuring it remains a crucial tool in safeguarding public trust and digital integrity.”
The system comprises two core components that support image, video, and audio detection. The core components are comprehensive detection and voice detection. The system scrutinizes various behavioral patterns, such as head angles, lip movements, and facial muscle changes, to determine the authenticity of the person being depicted. Additionally, the voice detection feature assesses content by analyzing frequency, time, and noise elements to detect manipulation. The detection process is efficient – typically taking between 5 to 10 minutes – after which the content is classified as “real” or “fake.” This system works for both videos and static images.
This deepfake detection system is designed based on DeepBrain AI’s generative AI avatar capabilities and deep learning-based AI human data that the company has accumulated internally. Unlike most existing detection models, which are created using Western-centric data, DeepBrain AI’s system increases detection accuracy by incorporating one million Korean data points and 130,000 Asian race data points.
Following the successful implementation of the system with the Korean National Police Agency, plans are underway to expand the deepfake detection offerings to other organizations and institutions. The goal is to reduce the social costs associated with false information and digital deception by providing tools that quickly verify the authenticity of content.
The solution is available as a Software-as-a-Service (SaaS) product through the company’s official website, allowing users to access the system without needing additional servers or equipment. For those requiring in-house deployment, an on-premise version is also available.
