Blockchain-based system aims to securely share medical AI data, modeled after Mayo Clinic
As medical artificial intelligence (AI) technology rapidly advances, the most significant concern is the risk of sensitive data leaks. Training medical AI requires massive amounts of medical data. There are fears that tech companies operating AI could access users’ health information or medical records, and that hacking could lead to the dissemination of such sensitive data.
In 2019, Google secretly partnered with Ascension, a large non-profit hospital chain in the U.S., to collect personal information and medical histories of at least 10 million people. Patients and medical staff were completely unaware that their data was being transferred to Google. The company planned to analyze this data to develop new medical software and AI.
Recently, AI companies have been protecting personal information by pseudonymizing sensitive medical data. However, the problem is that even pseudonymized data now carries a high risk of re-identification. Researchers at the Fraunhofer Institute in Germany pointed out that AI software can restore actual faces and identify individuals using only brain MRI scans. Current data anonymization measures are insufficient to fully protect personal information.
Hacking risks are also significant. A research team led by Professor Kim Young-hak from the Cardiology Department at Asan Medical Center in Seoul intentionally launched malicious attacks on medical AI models. The results showed that security measures were breached with approximately 81% probability, and the models exposed original data during responses up to 22% of the time.
To promote data sharing for medical AI while strengthening personal information protection, the government has launched a project to build a Korean-style Mayo Clinic platform. Mayo Clinic is a world-leading smart healthcare institution that operates an integrated medical AI and data platform. The government will invest 6 billion Korean won annually for three years to enable joint use of medical AI data. The plan involves using blockchain technology to anonymize personal information and fundamentally prevent leaks of sensitive data.
