Homomorphic encryption (FHE): the future of secret data

Technology companies, led by IBM, are working on the future of encryption with completely homomorphic encryption (FHE). This new form of encryption makes it possible to process information without the need for a key to unlock the data.

This piece is part of our ‘IT Explained’ series, where we explain important concepts and technologies behind products and innovations today in an understandable way. We regularly keep this piece up to date with the latest developments. Last updated: February 4, 2022.

Today we use classical encryption in different strengths according to different standards. However, the process is always the same: Data is made unrecognizable with an encryption key and can be unlocked with the same key. It works great for sending secure emails or private messages, for example, but when it comes to large datasets, the case becomes more sensitive.

When data is processed today, it must first be unlocked for processing. It opens up perspectives for theft. Those who first need to unlock this data for processing will gain insight into their content. With full homororph encryption, data can be processed securely without the need for keys.

First system

Why do not we do it in abundance today? Processing secure data is much more difficult because the content is unknown. The very first FHE system was built in 2009 by Craig Gentry, today an IBM Research specialist. It took 100 trillion times longer than that simple text to analyze.

The very first system in 2009 took 100 trillion times longer than plain text to analyze data.

The very first studies of this type of encryption date back to 1978, but it took until 2009 before computers were powerful enough to be able to prove their concept. Gentry used a grid-based cryptographic construction for the system that, unlike current encryption models, can withstand the future power of quantum computers.

Meanwhile, IBM has made significant progress, and other technology companies have embraced the technology to further develop it. Microsoft connected the technology to a neural network with deep learning in early 2016, and IBM rewrote the C ++ library in early 2018.

At the end of 2020, IBM was the first in the world to launch an FHE service package that education, security experts and prototype communities can get started on. This allows them to experiment with homororph encryption in an accessible way.


dr. Dave Braines, CTO Emerging Technology at IBM Research: “Today begins the era of homomorphic encryption becoming useful in certain situations. Depending on the dataset, it no longer takes 100 trillion times longer, but 1,000 times longer.” The best is actually because it can be up to 100,000 times longer, according to Braines. Much depends on the database and the available computing power. On the other hand, an existing system today must also decrypt and encrypt files at each calculation.

IBM does not expect homomorphic encryption to become the new standard for everything.

Homomorph encryption is an open source database that you can experiment with today. “We do not know what the final version will look like, the regulation surrounding it and the duration of the process,” Braines stressed. “High-level business APIs are already available today to speed up the process.”

IBM does not expect FHE to become the new standard for everything. Based on computing power, it costs way too much compared to classic encryption. Sectors such as healthcare, government, banking and insurance, machine learning and analytics benefit greatly from homomorph encryption. Privacy and secrecy are often essential, something that is not possible with the existing encryption method.


Braines brings up a few utility cases where the new form of encryption becomes important:

  • GPS location: You do not want to share your exact coordinates with a third party for commercial purposes. With the new encryption, the party only knows that you are nearby (yes or no) for a promo, but no details.
  • Financial details: Today, police can not inquire about details from a bank. This requires an official audit. With the new standard, it would no longer be necessary.
  • Bank and insurance: Today, they are often separated by separate encryption. Homomorphic encryption allows both to communicate without revealing details.
  • DNA information: Sensitive information can be processed by research laboratories without knowing the information.

In 2018, Numerai was one of the first parties to embrace the new form of encryption and to throw ‘expensive data’ online that anyone can use to train models. No one can see its contents, only the source and what price you can earn for processing the dataset.

Incidentally, IBM is not the only one involved in FHE. Intel announced in 2021 that they are working on a specific chip to speed up the process. The development of the chip will probably take several years and Microsoft is also part of the project.

quantum computer

IBM’s available toolkit allows organizations to experiment with homororph encryption. Braines: “We must all develop the competencies in relation to implementation. In addition, there must also be computer infrastructure somewhere. This can be local, but also in the cloud. You can use an unreliable platform for the calculation because there is no sensitive data available. The key always remains with the owner. “

When a quantum computer becomes powerful enough to break all of our current encryption methods, we need a solution.

Today, FHE still requires a lot of computing power. Work is underway on hardware acceleration, as is the case with the RSA standard today. This enables encryption technology to make great strides. Furthermore, we must not lose sight of developments in quantum computers.

Michael Osborne, Security Researcher at IBM Research: “When a quantum computer becomes powerful enough to break all of our current encryption methods, we need a solution. Some sectors such as automotive, aerospace and transportation are very slow moving sectors. New encryption standards are needed for “Lattice-based encryption methods are the future of which homororph encryption is a part.”

Leave a Comment