Deepfake detection made easy for everyone: SANDS Lab

[AI Revolution] (120) Deepfake detection made easy for everyone: SANDS Lab

Specializing in detecting and countering harmful uses of generative AI, including deepfakes and malware

As deepfake sex crimes emerge as a social problem, a Korean small and medium-sized company is attracting attention with a free service to help prevent them. That company is SANDS Lab, a security specialist in AI-based cyber threat intelligence (CTI).

Choi Min-jun, head of SANDS Lab’s AI technology development team, told Asia Economy on the 6th, “We are researching and developing technology in step with national efforts to prevent deepfake crime and fake news. We hope our deepfake detection technology will help many people as a contribution to society.”

The new service is FakeCheck, a deepfake detection solution. Users visit the website and upload a suspected image, and it analyzes authenticity and returns a result within seconds. The service focuses on its core function, eliminating registration and authentication so anyone can use it easily.

Choi explained, “FakeCheck uses multiple AI models to determine whether images of human faces have been manipulated. Because it is intended to prevent crime, it does not assess deepfake images of animals or objects. However, ongoing feedback and updates may expand detection targets and improve accuracy, and video manipulation detection could be supported in the future.”

Photo 1. Choi Min-jun, head of SANDS Lab’s AI technology development team

The service is free and open to everyone. SANDS Lab will cover all operating costs, including cloud services, as part of its social contribution efforts.

A major reason SANDS Lab could respond quickly to deepfake crime was its participation in an IITP national R&D project on generative AI. Together with 42Maru, KAIST, Logpresso and LG Uplus, it is conducting research under an information security core technology development project, taking responsibility for detection and response technologies for harmful uses of generative AI.

Choi said, “SANDS Lab and LG Uplus are leading the deepfake research. Only about four months have passed, so we cannot yet claim to have produced groundbreaking technology or solutions, but we intend to carry out the project well using SANDS Lab’s capabilities.”

SANDS Lab’s core specialty is cybersecurity, with CTX (Cyber Threat X) among its representative brands. CTX uses an AI engine to analyze files and support specific hash searches (MDS, SHA1 and SHA256), providing information on maliciousness, attack groups, target countries and industries, and MITRE ATT&CK techniques.

Photo 2. Choi Min-jun introduces CTX

Choi explained, “Ultimately, there comes a point when a person must decide that a file is malicious. CTX quickly provides information about when and where particular groups are targeting which industries and countries. By giving people convincing information, it speeds threat assessment and improves work efficiency.”

SANDS Lab also operates a dataset business using CTX. A dataset is a relevant collection of data for a particular task. Training AI requires high-quality datasets in a structured form. SANDS Lab’s datasets include information on diverse cyber threat groups and campaigns, helping AI algorithms respond to more realistic and varied situations.

Choi said, “Generative AI and dataset technologies are essential to further improving threat detection and response. We plan to accelerate global technology development through diverse collaborations, including with Korean generative AI specialists.”