ETRI develops innovative computing chip for homomorphic encryption

Korea’s Electronics and Telecommunications Research Institute (ETRI) announced on September 6 that it has developed a dedicated compute accelerator chip capable of handling high-speed homomorphic encryption, which is currently gaining traction. attracted much attention as a next-generation cryptographic technology.

Unlike conventional password encryption, homomorphic encryption is a technology that directly computes encrypted data without re-resolving. There is no need to use the private key to decrypt the encrypted data, so there is no risk of the secret key or the original data being exposed. It is also quantum resistant, meaning it cannot be cracked using quantum computing.

ETRI researchers believe that once the technology is commercialized, it will be used in data center servers for cloud services and artificial intelligence semiconductors to ensure privacy. Research on chips and software development platforms for homomorphic cryptographic accelerators is also expected to gain momentum.

This technology is expected to enable data to be encrypted and directly applied to various converged services where concurrent security, statistics and artificial intelligence applications are required to be able to apply to the defense, public, medical, financial and industrial sectors, among many others. It can be suitable for cloud computing and database management system (DBMS) companies, great companies, server-mounted homomorphic cryptographic accelerator developers, translation companies, etc. data security services and companies that use sensitive data such as public and defense data.

This ETRI-developed fully homomorphic hardware compute acceleration chip is a dedicated hardware compute processing unit that conforms to the unique characteristics of homomorphic coding in which ciphertext is represented. by a polynomial of tens of thousands of degrees or more with a coefficient of thousands of bits.

They perform fast operations such as addition, subtraction, multiplication, and division among higher-order polynomials while the data remains encrypted. The principle is that encrypted data is directly computed and only the result can be decrypted to reveal the data. No one can see the original data so data privacy is guaranteed.

When the data is converted to ciphertext, it becomes a polynomial tens of thousands of times larger, increasing the amount of computation exponentially. Therefore, the use of homomorphic encoding is quite a challenge because it is not suitable for computation with a central processing unit (CPU) or a graphics processing unit (GPU) that processes 64-bit data about structural face. a lot of time.

This research project solves these problems and provides the core technology for homomorphic processing units (HPUs) that can use homomorphic encryption, the researchers explain.

Based on this technology, the researchers plan to deploy and commercialize a system on a semiconductor chip (SoC) dedicated to homomorphic encryption in the future. Furthermore, they will develop a fully homomorphic hardware accelerator scaling platform that can infer a homomorphic encoded video into a real-time homomorphic encoded artificial intelligence model with AI model supported by homomorphic cryptographic SoC.