#South Korea #Seoul #Private AI #Desilo #Fully Homomorphic Encryption

DESILO's Revolutionary GL Scheme

In a momentous announcement made on May 14, 2026, DESILO, a leader in Fully Homomorphic Encryption (FHE) technologies, revealed that two research papers on its 5th-generation homomorphic encryption scheme, known as 'GL', have been accepted at the prestigious Crypto 2026 conference. This event, renowned in the cryptography community, underscores the significant advancements made by the company in making data processing safer and more efficient.

The acceptance of the two papers, which focus on the core scheme of GL and the bootstrapping process allowing unlimited computation in homomorphic encryption, marks an exceptional achievement in academic recognition. Traditionally, having two closely related papers accepted simultaneously at a top-tier conference is rare, highlighting the depth and innovation represented in DESILO's work.

Co-authored by distinguished researchers including DESILO's own Yongwoo Lee, Eunmo Yang, Youngmin Lee, and Hyojun Kim, alongside FHE pioneer Craig Gentry, who first developed a fully homomorphic encryption scheme in 2009 and received the Gödel Prize, the papers offer groundbreaking insights into the practical applications of FHE technology. Craig's involvement, especially with his historical contributions to cryptography, adds further credibility and prestige to DESILO's advancements.

The significance of the GL scheme cannot be overstated. In a world where the integration of artificial intelligence (AI) into sensitive data processing is rapidly becoming essential, the need for secure methods of operation has never been more crucial. DESILO's GL scheme accelerates data security by enabling computations directly on encrypted data, ensuring that sensitive information remains protected from unauthorized access.

One of the core innovations of the GL scheme is its ability to dramatically enhance processing speeds over previous generations of FHE technology. By focusing on optimizing matrix operations, which are fundamental to AI and machine learning, the GL scheme achieves performance improvements on the order of several hundred times. Such enhancements pave the way for the commercialization of 'Private AI', a sector emphasizing the use of AI while safeguarding raw data—addressing one of the main challenges faced in this rapidly advancing field.

Following the publication of these papers, DESILO is not merely resting on its laurels. The company is actively embedding the GL scheme into its commercial FHE library, DESILO FHE, which aims to solidify the technology's practical applications. As Yongwoo Lee, the Chief Scientist at DESILO, emphasized, the simultaneous acceptance of both papers showcases GL as a comprehensive system ready for real-world implementation rather than a theoretical concept.

In conclusion, DESILO's achievements with the GL scheme represent a paradigm shift in the realm of encryption technologies, particularly in their applications to AI. As the company continues to roll out its innovations, it stands poised to not only lead the field of secure AI processing but also redefine the standards of privacy in the digital era. This recognition at the IACR Crypto conference signals a bright future for both DESILO and the broader field of privacy-enhancing technologies, ensuring that as AI continues to develop, the protection of sensitive data remains paramount.