Professor Kim Young-sik of Chosun University Co-develops Encryption Technology to Ensure Data Privacy

Professor Kim Young-sik, Chosun University. Photo by Chosun University

[Asia Economy Honam Reporting Headquarters, Reporter Cho Hyung-joo] Professor Kim Young-sik of the Department of Information and Communication Engineering at Chosun University, together with a research team from Seoul National University, has developed a high-precision encryption technology and presented a paper titled “High-precision bootstrapping of RNS-CKKS homomorphic encryption using optimal minimax polynomial approximation and inverse sine function” at Eurocrypt 2021, one of the world’s top two academic conferences in the field of cryptography.

According to Chosun University on the 1st, Eurocrypt is one of the top-tier academic conferences designated by the National Research Foundation of Korea, and this year, the paper involving Professor Kim was the only one presented from Korea.

The paper, published on October 18, dealt with high-precision bootstrapping, a core technology for applying CKKS homomorphic encryption?which allows computations on encrypted data?to artificial intelligence.

When this technology is utilized, it increases the accuracy of the bootstrapping technique by more than a thousand times, enabling the practical use of homomorphic encryption and the development of AI technologies that guarantee data privacy.

Currently, high-quality data is essential for AI technology development, but since data often contains sensitive information such as personal data, there are many restrictions due to privacy concerns and legal regulations.

However, by using homomorphic encryption, sensitive information can be encrypted and AI computations can be performed on the encrypted data without decryption, protecting privacy while obtaining the desired results, thus providing a fundamental solution to this problem.

For the use of homomorphic encryption, performing computations is essential, but as computations proceed, the available computational resources decrease, so bootstrapping techniques must replenish these resources for homomorphic encryption.

However, existing bootstrapping methods had the drawback of introducing significant errors during this process, making it impossible to continue computations.

In other words, the total amount of computation possible with homomorphic encryption was very limited, but this new technology allows computations to be performed as much as desired.

Deep neural networks, which are at the center of today’s AI innovation, involve extensive computations, and previously, applying homomorphic encryption to AI computations was impossible; however, this result now makes it possible to apply homomorphic encryption to deep neural network computations.

The research team actually implemented this technology and, in collaboration with Samsung Advanced Institute of Technology, demonstrated for the first time in the world that encrypted images could be successfully classified by AI using the standard deep neural network AI model ResNet-20. They are also conducting research on training encrypted AI models.

Currently, homomorphic encryption is attracting attention as one of the core cryptographic technologies of the future, with various research groups in the United States and Europe competing for its practical application.

This paper was co-researched by Professor Kim Young-sik of Chosun University, Professor Noh Jong-seon of Seoul National University, and researchers Lee Jun-woo, Lee Eun-sang, and Lee Yong-woo (currently affiliated with Samsung Advanced Institute of Technology), supported by Samsung Electronics Advanced Institute of Technology.

Professor Kim Young-sik said, “This result marks a groundbreaking turning point for privacy-preserving AI technology that applies AI systems directly to encrypted data,” and added, “Going forward, our research team will focus on practical research through high-speed implementation of homomorphic encryption.”

Meanwhile, in April, Professor Kim Young-sik and the Chosun University research team began a project to secure foundational technologies for post-quantum cryptographic algorithms and security verification, selected for the core information security technology development project supported by the Institute for Information & Communications Technology Planning & Evaluation (IITP).

The research team will receive funding for about 3 years and 9 months starting from April 2021 under the project titled “Development of high-efficiency PQC security and performance verification technology for low-spec devices.”