Original Author: Mustafa Hourani
Original Translation: IOSG Team
Preface
Introducing Fully Homomorphic Encryption (FHE): Exploring its exciting applications, limitations, and driving its recent popularity.
When I (Mustafa) first heard about "Fully Homomorphic Encryption" (FHE), I thought of the tendency in the blockchain industry to give long names to popular concepts. Over the years, we have encountered many buzzwords that have caused a sensation in the industry, with the most recent being "Zero-Knowledge Proof" (ZKP).
After conducting some research and exploration into new companies using FHE to build products, I noticed a horizon filled with brilliant new tools. In the coming months and years, FHE may become the next big technology to sweep the industry, much like ZKP. Companies are leveraging the latest advances in cryptography and cloud computing to pave the way for a powerful, data-privacy-protecting future. The question is not whether we can achieve it, but when we can achieve it, and I believe FHE may be the key catalyst driving forward data privacy and ownership.
"FHE is the holy grail of cryptography. Over time, FHE will reshape the structure of all computations, whether it's web2 or web3."
What is Homomorphic
The term "Homomorphism" refers to the meaning of "homomorphic." Tracing its roots, homomorphism originated in mathematics, defined as a mapping between two similar algebraic structures that preserves a core component between them.
If, like me, you prefer a practical definition, the underlying mathematical principle is that two groups do not need to be exactly the same to have the same core properties. For example, imagine two boxes containing fruits, with each box corresponding to a different group:
Box A contains small fruits.
Box B contains large fruits.
Despite the different sizes of the fruits, juicing small apples and oranges together in Box A can produce the same mixed fruit juice flavor as juicing large apples and oranges together in Box B. Juicing to produce the same flavor is similar to maintaining a core component between the two boxes. Assuming the same flavor is our main concern, it doesn't matter from which box we juice the fruits, as the quantity of juice is not our focus. In the important aspect (flavor), these two groups are equivalent, so their differences (size and quantity) do not affect our definition of their main function, which is to produce a specific fruit juice flavor.
By analogy with homomorphism, we capture its two main features:
Mapping: We establish a connection between the fruits, where each small fruit in Box A corresponds to a larger version in Box B. Therefore, a small apple in Box A corresponds to a large apple in Box B, and so on.
Operation preservation: If juicing two small fruits in Box A produces a specific flavor, then juicing their larger versions in Box B should also produce the same flavor. Despite the differences in the size and quantity of the juice obtained, the "flavor feature" is preserved.
What is Fully Homomorphic Encryption
In connection with the central topic of this article, Fully Homomorphic Encryption (FHE) is a specific data encryption method that allows people to perform calculations on encrypted data without exposing the original data. In theory, the analysis and calculations performed on encrypted data should produce the same results as those performed on the original data. Through FHE, we establish a one-to-one correspondence between the data in the encrypted data set and the data in the original data set. In this case, the preservation of the core component is the ability to perform any calculation on the data in any data set and produce the same results.
In this regard, many companies have taken precautions to protect user data and maintain differential privacy. Companies rarely store raw, unencrypted data in the cloud or their databases. Therefore, even if attackers control a company's servers, they would still need to bypass the encryption to read and access the data. However, when data is only encrypted and not used, it is not interesting. When companies want to analyze data to gain valuable insights, they have no choice but to decrypt the data. When data is decrypted, it becomes vulnerable. However, with end-to-end encryption, FHE becomes very useful because we no longer need to decrypt the data to analyze it; this is just the tip of the iceberg of possibilities.
A key consideration is whether companies should be allowed to read and store our personal information from the outset. Many people's standard response to this is that companies need to see our data in order to provide us with better services.
If YouTube does not store data such as my viewing and search history, the algorithm cannot fully realize its potential to show me videos I am interested in. Therefore, many people believe that the balance between data privacy and obtaining better services is worthwhile. However, with FHE, we no longer need to make this trade-off. Companies like YouTube can train their algorithms on encrypted data and produce the same results for end users without infringing on data privacy. Specifically, they can homomorphically encrypt information such as my viewing and search history, analyze it without viewing it, and then show me videos I am interested in based on the analysis.
FHE is an important step towards establishing a future where our data is no longer a valuable commodity freely given to organizations.
Applications of Fully Homomorphic Encryption
If fully homomorphic encryption (FHE) is correctly applied, it will be a breakthrough for all industries that store user data. We are witnessing a technology that could change our overall attitude towards data privacy and the acceptable limits of intrusion by companies.
Let's first explore how FHE could transform data practices in the healthcare industry. Many hospitals store private records of patients in their databases, which must be kept confidential for ethical and legal reasons. However, this information is very valuable to external medical researchers, who can analyze this data to derive important insights about diseases and potential treatment methods. A major obstacle slowing down research progress is maintaining the complete confidentiality of patient data when outsourcing it to researchers. There are many methods to anonymize or pseudonymize patient records, but these methods are not perfect, either potentially revealing too much information about an individual, making them identifiable, or not revealing enough case information to obtain accurate insights about diseases.
Through FHE, hospitals can homomorphically encrypt patient data, making it easier to protect patient privacy in the cloud. Medical researchers can perform calculations and run analysis functions on encrypted data without compromising patient privacy. Since there is a one-to-one mapping between the encrypted data and the original data, the results obtained from the encrypted data set provide real insights that can be applied to actual cases. FHE can rapidly drive progress in the healthcare industry.
Fully Homomorphic Encryption (FHE) has another exciting application in artificial intelligence (AI) training that is worth paying attention to. Currently, the AI field faces privacy issues, which hinder companies' ability to access the large and diverse datasets crucial for refining AI algorithms. Companies training AI must choose between using limited public datasets, paying a significant amount of money to purchase private datasets, or creating datasets, which is particularly challenging for smaller companies with fewer users. FHE should be able to address the privacy issues that prevent many dataset providers from entering this market. Therefore, improvements in FHE are likely to lead to an increase in the number of datasets available for AI training. This will make AI training more economically viable and refined, given the increased diversity of available datasets.
With FHE, companies can train machine learning models on encrypted data without revealing the original data. This means that data owners can securely share their encrypted data without worrying about privacy breaches or data misuse. At the same time, AI model trainers can improve their algorithms using more diverse and comprehensive datasets, which may not be possible without FHE. Therefore, fully homomorphic encryption not only enhances data security but also broadens the possibilities for AI research and development, making the application of AI technology more widespread and efficient.
Past Limitations of Fully Homomorphic Encryption
Although Fully Homomorphic Encryption (FHE) does promise to change contemporary big data, why haven't we seen more practical applications yet?
While FHE has been a topic of discussion and research for many years, the reality is that implementing FHE in practice is extremely challenging. The core challenge lies in the computational power required to perform FHE. A fully homomorphically secure dataset should produce analysis results identical to its original data form. This is a challenging feat that requires a significant amount of computational speed and power, much of which is impractical to implement on existing computers. Operations that typically take seconds on raw data may take hours or even days on homomorphically encrypted datasets. This computational challenge creates a self-perpetuating cycle, with many engineers delaying the undertaking of FHE projects, thus slowing its development and limiting the full realization of its advantages.
Engineers face a specific computational problem with FHE, which is dealing with "noise errors." When performing calculations on homomorphically encrypted datasets, many engineers encounter additional noise or errors with each computation. While this may be tolerable when only a few calculations are needed, after multiple analyses, the noise may become so pronounced that the original data becomes difficult to understand. The data is almost lost.
Why Now
Just as generative AI was once considered limited and primitive until it became mainstream, Fully Homomorphic Encryption (FHE) is progressing along a similar trajectory. Many industry leaders, even those beyond the blockchain field, have gathered to organize extensive research and development on FHE. This has led to several recent industry developments, driving the remarkable narrative of technological advancement.
DPRIVE Project
In March 2021, Microsoft, Intel, and the Defense Advanced Research Projects Agency (DARPA) agreed to launch a multi-year program to accelerate the development of Fully Homomorphic Encryption (FHE). This program, called "Data Protection in Virtual Environments" (DPRIVE), marks a significant advancement in FHE. It demonstrates the collaboration of two industry giants focused on cloud computing and computer hardware to address data privacy issues. They initiated this program to build computers and software capable of managing the computational speed required for FHE and to establish precise guidelines for the accurate implementation of FHE, preventing data leaks caused by misuse.
As part of the DPRIVE program, engineers have taken on the task of reducing noise to maintain the level of original data when performing computations on homomorphically encrypted datasets, thus mitigating the previously mentioned "noise errors." A promising solution is the design of Large Arithmetic Word Sizes (LAWS) data representation. While traditional central processing units (CPUs) typically use 64-bit words, engineers are developing new hardware capable of handling 1024-bit or larger words for LAWS. This approach is effective because research has shown that longer words directly impact the signal-to-noise ratio. In simple terms, longer words in FHE produce less noise with each additional computation, allowing for more computations to be performed until reaching the data loss threshold. By building new hardware to address these challenges, engineers participating in the DPRIVE program have significantly reduced the computational load required for FHE.
To accelerate computations and approach the goal of increasing FHE speed by 100,000 times, the DPRIVE team embarked on a continuous journey to design a new data processing system that surpasses the capabilities of conventional processing and graphics units. They developed a new Multiple Instruction Multiple Data (MIMD) system that can simultaneously handle multiple instructions and datasets. MIMD is akin to constructing a new high-speed highway rather than using existing inadequate roads to accommodate the traffic required for fast, real-time FHE computations.
An interesting aspect of the DPRIVE program is its extensive use of "parallelism" in computer mathematical computations. This allows developers to perform multiple large number calculations simultaneously. You can think of parallelism as deploying a group of mathematicians to work on different parts of a massive mathematical problem simultaneously, rather than having them work on their tasks one after another. While performing multiple computations simultaneously helps solve problems quickly, the computer must undergo air cooling to prevent overheating.
In September 2022, over a year and a half after the launch of the program, Microsoft, Intel, and DARPA announced that they had successfully completed the first phase of the DPRIVE program. They are currently proceeding with the second phase of DPRIVE.
SDKs and Open Source Libraries
With numerous large companies pioneering in the field of Fully Homomorphic Encryption (FHE), the availability of software development kits (SDKs) and open-source libraries has increased significantly, allowing developers to build on each other's work.
Microsoft announced the release of Microsoft SEAL, an open-source library that provides developers with tools to perform homomorphic encryption on datasets. This enables a wider range of developers to explore FHE, democratizing access to end-to-end encryption and computational services. The library provides examples of homomorphic encryption programs and detailed comments to guide developers in the correct and secure usage.
Intel also introduced its own homomorphic encryption toolkit, providing developers with tools to implement homomorphic encryption more quickly in the cloud. Intel designed this toolkit with flexibility to ensure compatibility with the latest advancements in data processing and computation. It includes functions customized for lattice cryptography, seamless integration with Microsoft SEAL, samples of homomorphic encryption schemes, and technical documentation to guide users.
Google's Private Join and Compute open-source library provides developers with tools for multi-party computation (MPC). This computational method allows parties to combine their different datasets to obtain shared insights without exposing their original data to each other. Private Join and Compute combines cryptographic techniques from FHE with Private Set Intersection (PSI) to optimize data privacy practices. PSI is another cryptographic method that allows parties with different datasets to identify common elements or data points without revealing their data. Google's approach to advancing data privacy not only focuses on FHE but also integrates FHE with other influential data practices, prioritizing a broader concept of MPC.
It is worth noting that the availability of reputable open-source libraries for FHE is on the rise. However, this becomes even more noteworthy when observing well-known companies experimenting with these libraries in their operations. In April 2021, as a prominent global technology entity in stock exchanges and capital markets, Nasdaq incorporated FHE into its operations. Nasdaq utilized Intel's FHE tools and high-speed processors to combat financial crimes through anti-money laundering efforts and fraud detection. This was achieved by using homomorphic encryption to identify valuable insights and potential illicit activities in datasets containing sensitive information.
Recent Capital Raises
In addition to the research and development efforts mentioned earlier, several other companies have recently secured significant funding for initiatives focused on Fully Homomorphic Encryption (FHE).
Cornami is a large technology company known for developing scalable cloud computing technology specifically designed for homomorphic encryption. They have been involved in many efforts to create computing systems that are more efficient in supporting FHE than traditional CPUs. They also lead initiatives aimed at protecting encrypted data from quantum computing threats. In May 2022, Cornami announced the successful completion of a Series C funding round, raising $68 million led by SoftBank, bringing their total capital to $150 million.
Zama is another company in the blockchain industry that is building open-source homomorphic encryption tools, allowing developers to build exciting applications using FHE, blockchain, and AI. Zama has built a Fully Homomorphic Ethereum Virtual Machine (fhEVM) as part of their product offering. This smart contract protocol keeps on-chain transaction data encrypted during processing. Developers exploring various applications of the Zama library have been impressed by its performance, even in complex use cases. In February 2022, Zama successfully completed a $42 million Series A funding round led by Protocol Labs, raising their total capital to $50 million.
Fhenix is also an emerging project that is bringing FHE into blockchain. Their goal is to extend the application of FHE beyond privacy payments, opening the door to exciting use cases for FHE in decentralized finance (DeFi), bridging, governance voting, and Web3 games. In September 2023, Fhenix announced the completion of a $7 million seed round led by Multicoin Capital and Collider Ventures.
What's Next
For years, Fully Homomorphic Encryption (FHE) has been a promising idea for providing robust end-to-end encryption, heralding a powerful future for data privacy. Recent developments are beginning to transform FHE from a theoretical dream into practical applications. While various companies are competing to be the first to implement a powerful, fully functional version of FHE, many companies are collaborating to navigate the complexity of this powerful technology. This spirit of collaboration is evident through their implementation of various cross-team projects and the development of open-source libraries integrated with other libraries.
Based on my findings, the discussions surrounding FHE seem to have far-reaching implications. In the coming weeks, I am excited to delve deeper and share more insights into my research on FHE. Specifically, I am eager to explore more about the following topics:
Emerging applications of FHE:
Interaction between Zero-Knowledge Proofs (ZKPs) and FHE.
Integrating FHE with Private Set Intersection (PSI) to advance secure Multi-Party Computation (MPC).
New companies like Zama and Fhenix leading the development in the FHE field.
Reference:
Arampatzis, Anastasios. “Latest Developments in Homomorphic Encryption.” Venafi, 1 Feb. 2022, venafi.com/blog/what-are-latest-developments-homomorphic-encryption-ask-experts/.
Arampatzis, Anastasios. “What Is Homomorphic Encryption & How Is It Used.” Venafi, 28 Apr. 2023, venafi.com/blog/homomorphic-encryption-what-it-and-how-it-used/.
“Building Hardware to Enable Continuous Data Protections.” DARPA, 2 Mar. 2020,www.darpa.mil/news-events/2020-03-02.
Cristobal, Samuel. “Fully Homomorphic Encryption: The Holy Grail of Cryptography.” Datascience.Aero, 7 Jan. 2021, datascience.aero/fully-homomorphic-encryption-the-holy-grail-of-cryptography/.
“Homomorphic Encryption: What Is It, and Why Does It Matter?” Internet Society, 9 Mar. 2023,www.internetsociety.org/resources/doc/2023/homomorphic-encryption/.
Hunt, James. “FHENIX Raises $7 Million in Seed Round Led by Multicoin Capital.” The Block, The Block, 26 Sept. 2023,www.theblock.co/post/252931/fhenix-seed-multicoin-capital.
“Intel® Homomorphic Encryption Toolkit.” Intel, accessed 8 Oct. 2023,www.intel.com/content/www/us/en/developer/tools/homomorphic-encryption/overview.html#gs.fu55im.
“Intel to Collaborate With Microsoft on DARPA PROGRAM.” Intel, 8 Mar. 2021,www.intel.com/content/www/us/en/newsroom/news/intel-collaborate-microsoft-darpa-program.html#gs.ftusxq.
“Intel Xeon Advances NASDAQ’s Homomorphic Encryption R&D.” Intel, 6 Apr. 2021, www.intel.com/content/www/us/en/newsroom/news/xeon-advances-nasdaqs-homomorphic-encryption-rd.html#gs.6mpgme.
Johnson, Rick. “Intel Completes DARPA DPRIVE Phase One Milestone for a Fully Homomorphic Encryption Platform.” Intel, 14 Sept. 2022, community.intel.com/t5/Blogs/Products-and-Solutions/HPC/Intel-Completes-DARPA-DPRIVE-Phase-One-Milestone-for-a-Fully/post/1411021.
“Microsoft Seal: Fast and Easy-to-Use Homomorphic Encryption Library.” Microsoft Research, 4 Jan. 2023, www.microsoft.com/en-us/research/project/microsoft-seal/.
Paillier, Dr. Pascal. “Fully Homomorphic Encryption: The Holy Grail of Cryptography.” Business Age, 9 Mar. 2023, www.businessage.com/post/fully-homomorphic-encryption-the-holy-grail-of-cryptography.
Samani, Kyle. “The Dawn of On-Chain FHE.” Multicoin Capital, 26 Sept. 2023, multicoin.capital/2023/09/26/the-dawn-of-on-chain-fhe/.
Walker, Amanda, et al. “Helping Organizations Do More Without Collecting More Data.” Google Online Security Blog, 19 June 2019, security.googleblog.com/2019/06/helping-organizations-do-more-without-collecting-more-data.html.
“What Is Fully Homomorphic Encryption?” Inpher, 11 Apr. 2021, inpher.io/technology/what-is-fully-homomorphic-encryption/.
White, Matt. “A Brief History of Generative AI.” Medium, 8 July 2023, matthewdwhite.medium.com/a-brief-history-of-generative-ai-cb1837e67106#:~:text=Although%20most%20people%20will%20admit,of%20Stability%20AI’s%20Stable%20Diffusion.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
