MIT Researchers Describe a Somewhat Homomorphic Encryption Scheme

Summary:

What MIT Built

Homomorphic encryption lets someone run computations on encrypted data without ever seeing the data itself. A hospital could, for example, use a cloud service to run AI analysis on patient records while keeping them private. The catch is that existing schemes are so computationally heavy that they are often impractical. In March 2025, MIT researchers described a new, simpler theoretical way to build such schemes.

How the New Scheme Works

The researchers combined two simple cryptographic tools. They started with a linear homomorphic encryption scheme, which can only add encrypted values, and added one theoretical assumption that “lifts” it into a somewhat homomorphic scheme able to handle more complex functions.

Why Noise Is the Problem

Homomorphic schemes add noise to a message to encrypt it, and that noise grows as operations run. If one computes for too long, the noise can drown out the message. The new scheme limits computations to bounded polynomials, which allow many additions but only a few multiplications, keeping noise under control. Each piece of data is encrypted into a matrix that provably hides it, and computations are done by adding or multiplying those matrices.

What It Could Mean for AI

The long-term goal is private AI. “The dream is that you type your ChatGPT prompt, encrypt it, send the encrypted message to ChatGPT, and then it can produce outputs for you without ever seeing what you are asking it,” said Henry Corrigan-Gibbs, an associate professor in MIT’s Department of Electrical Engineering and Computer Science (EECS) and a co-author of the paper.

Potential Applications

In the near term, the somewhat homomorphic approach could support private database lookups and private statistical analysis. Its simpler math could make it efficient enough for a wider range of real-world uses than today’s heavyweight schemes.

The Research Team

Corrigan-Gibbs’s co-authors are EECS graduate student Alexandra Henzinger, Professor Yael Kalai and Professor Vinod Vaikuntanathan, a principal investigator in MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL). The research will be presented at the International Conference on the Theory and Applications of Cryptographic Techniques.

Limits and Next Steps

The work is theoretical. The next step is making the scheme fast enough to run on modern hardware.

How Far Off Is It?

“We haven’t spent 10 years trying to optimize this scheme yet, so we don’t know how efficient it could get,” Henzinger said. The team also hopes to extend the technique to support more complex operations, possibly moving toward a new approach to fully homomorphic encryption.

The Bigger Picture

Homomorphic encryption has been studied since the 1970s, but the first scheme wasn’t built until 2009. Progress has been slow because security and flexibility pull in opposite directions.

This work adds a new, simpler route. It is not ready for products yet, but it widens the options for protecting data while it is being used.

Fred Templeton, CISA, CASP, SEC+

Fred Templeton is a practicing Information Systems Auditor in the Washington DC area. Fred works as a government contractor and uses his skills in cyber security to make our country's information systems safer from cyber threats. Fred holds a master's degree in cybersecurity and is currently working on his PHD in Information Systems.